KAP & UAV Exposure Control Intervalometer Script

kona

Thank you for creating such a great script. It super charged my S110.
I was wondering if it's possible to adjust a few settings to get a sharper image from my camera. I'm flying a winged plane between 50-100 meter grid missions. It seems to me that the adjustments are off a bit. I followed the instructions for the camera and set the Kap 3.6 settings to burst, pixhawk and checked focus on infinity. All other settings are default. Any suggestions to get a sharper image from my Canon S110 would be greatly appreciated. I attached a photo and kap.log file. Thank you again.

KAP.log

487.68KB

waterwingz

kona
It seems to me that the adjustments are off a bit. I followed the instructions for the camera and set the Kap 3.6 settings to burst, pixhawk and checked focus on infinity. All other settings are default. Any suggestions to get a sharper image from my Canon S110 would be greatly appreciated. I attached a photo and kap.log file.

Thanks for including the log file. Actually, your one posted image does not look too bad to me. Are some of the others you took better than the others or are they all about the same?

A couple of things I see in the log file (which could be bugs in the log file I suppose)

According to the log file, you have "Focus at Infinity" disabled rather than enabled.
You have Continuous AF enabled in the Canon menu. Please disable it.
Your target Tv value is 1/1000 and your max tv value is 1/2000. According to the log there is lots of light available and it's only using ISO100 so crank the target Tv to 1/2000 (or even 1/5000) and the max TV to something higher than that.

Unrelated to camera settings, your vibration mount can make a big difference. What are you doing there ?

Finally, here's a note about Canon cameras - and specifically the S90-S120 series : Let's Talk Canon Cameras

Note : I've asked for this thread to be merged into the main kap_uav.lua thread

edit : typo - disabled rather than disabled should read disabled rather than enabled

kona

Thank you so much for the fast reply. Most of my photos look like the one I posted. Most of the photos are about the same as what I posted. I would like to get just a little sharper image.

I will update the settings you suggested. I wonder why the focus on infinity shows disabled when there is a dot next to it. You are talking about the setting in the KAP script, right?

I have the camera mounted to a foam block that is velcroed to coroplast. The camera also shoots through a 1- 1/2 square piece of double strength clear glass (approx 1/8" thick). This glass is mounted between the camera lens and the bottom of the plane. Perhaps I need to redesign my camera mount to help cut down on vibration.

I really appreciate your suggestions. I look forward to taking more photos in the morning. Thanks again.

waterwingz

kona
I wonder why the focus on infinity shows disabled when there is a dot next to it. You are talking about the setting in the KAP script, right?

My mistake - your setting are correct. The log file you posted has data from 27 different runs of the script. In the first 16 of those, you had "Focus @ Infinity" disabled - in the last 11 runs you had it enabled.

kona

I'm still having trouble getting the script to be in focus with my Canon S110. If I use autofocus the photos are good. If I turn off autofocus and set the radial button in the script to focus on infinity the photos are slightly blurred. See attachments. I would leave autofocus on but I get less photos during the mission. Maybe I'm missing a step? I already tried going into the camera settings and manually set the camera to infinity prior to flying but end up with the same slightly blurred photos.

Any help would be greatly appreciated.

Thank you.

kona

Forgot to include the KAP.LOG

KAP.log

332.99KB

waterwingz

kona
Maybe I'm missing a step?

I don't think you have missed anything, but see my last comment below.

It is my opinion that we are dealing with a fundamental limitation with trying to use MF on Powershot cameras. I've suggested my theory about this before - your most recent observations seem to add confirmation.

I already tried going into the camera settings and manually set the camera to infinity prior to flying but end up with the same slightly blurred photos.

This is the key. It's not related to the script settings if using Canon MF mode also has the same problem.

Simply put, the distance values (especially infinity) that you enter in Canon MF mode (or via a CHDK override) don't really work well. The camera focus mechanism is not well calibrated, if it's even calibrated at all. So when you give the camera a focus distance, it moves the lens to a fixed position that it looks up in a table and interpolates for. If this position is not correct, then the focus will not be good. In the case of your pictures, I would suggest the focus is close but not perfect, which is consistent with my theory.

So why does autofocus work so much better? Well, in autofocus mode the camera does not care about or use distance. It simply moves the lens focus mechanism until it finds the sharpest focus (based on "crispness" of observed image edges at the pixel level). The selected focus is then at the best position, but distance is not known or needed.

I would leave autofocus on but I get less photos during the mission.

So I see you are using the new pixhawk mode? I have not completed the documentation on that so I'm curious about how you have it setup. Thank for posting the log file.

You should be able to leave the camera in autofocus mode when setup for pixhawk shooting and only the first shot in a shooting sequence will focus. The rest will use the focus & exposure settings from the first shot and you should be able to get about two shots per second that way.

Also, from the log, I see that you have the camera set to Canon continuous shooting mode (which is fine) but I wonder what happens if you leave it in single shot mode, set the script's shot interval to zero in the script, and let the script manage shooting continuously? You should get the same frame rate but the camera might handle focus resetting differently ?

Naccio

Hey WW! Quick question: When using the new Pixhawk mode, I just have to connect the camera to one of the RC out channels, right? If that is the case, with a few script modifications (amount of PWM pulses needed for each action) I could connect it directly to the RC receiver and control the camera just with my RC transmitter, bypassing the pixhawk? That would mean the script would be ready to be used by any kind of flight controller, as they all have at least four RC PWM outputs...

waterwingz

Naccio
Hey WW! Quick question: When using the new Pixhawk mode, I just have to connect the camera to one of the RC out channels, right?

Correct.

If that is the case, with a few script modifications (amount of PWM pulses needed for each action) I could connect it directly to the RC receiver and control the camera just with my RC transmitter, bypassing the pixhawk?

That will not work - good try though.

The pixhawk setup that we discovered somewhat by trial and error generates pulses in the range of about 4 mSec to 20 mSec, which CHDK can discriminate if high precision USB timing is enabled. The servo output of your RC radio generated pulses shorter than 2 mSec, too fast for CHDK to read.

That would mean the script would be ready to be used by any kind of flight controller, as they all have at least four RC PWM outputs...

I did some experimenting with this to try and get CHDK to read servo outputs directly. I was able to almost discriminate three stick positions - down, center, up - but it was not extremely reliable. Doing some averaging over multiple pulses helped with that but it was very processor intensive and would likely not have worked well with low end cameras. I'll see if I can find the post where I reported the results.

c_joerg

Iâ€™m not sure if this is helpful:

I have also an S110 and manual focus setting does not work very well.

Autofocus works very well. Only think which I notice: I got absolute unrealistic values with get_focus, like object distance is 4m, I got 1.2m with get_focus. This is not happens on my G1x. These values are more realistic.

Also when I look in the EXIF Data on my S110 pictures with infinity objects. All pictures are sharp. But I donâ€™t find infinity values in EXIF (May be 1 in 200 pictures).

This is totally different on my IXUS500HS, G1X and SX230HS. Here I find a lot of infinity values in EXIF.
Setting manual focus (like hyperfocal distance) works very well on G1X and SX230HS, but not on S110.

Naccio

waterwingz
The pixhawk setup that we discovered somewhat by trial and error generates pulses in the range of about 4 mSec to 20 mSec, which CHDK can discriminate if high precision USB timing is enabled. The servo output of your RC radio generated pulses shorter than 2 mSec, too fast for CHDK to read.

Ok, so you modify the pixhawk settings to produce out of spec RC PWM values... Did anybody check with the Pixhawk developers if that could have any repercussions on the pixhawk hardware (physical damage, I suppose there is very low chance) or software (sofware crash midflight :'()?

waterwingz

Naccio
Did anybody check with the Pixhawk developers if that could have any repercussions on the pixhawk hardware

No.

Although I did read through the source code - it's open source and available for anyone to download. From what I could see, using the values we "discovered" should not cause any issues.

One of the kap_uav.lua users was trying unsuccessfully to engage in a dialog about the application with the pixhawk team. If you have a contact there feel free to reach out to them?

waterwingz

c_joerg
I have also an S110 and manual focus setting does not work very well.

And yet manual focus seems to work really well on my S100.

Other people I have talked to who own several S100's report at MF works well on some but not others.

Like I said earlier, I believe it's something that is not well calibrated. If you get lucky, then the MF value you specify sets the lens to the right position. If you don't get lucky, it does not.

It's possible that cameras in the more expensive G series are actually calibrated - or the lens mechanism does not need calibration.

Naccio

waterwingz
Although I did read through the source code - it's open source and available for anyone to download. From what I could see, using the values we "discovered" should not cause any issues.

I tried to follow the code but couldn't get past the HAL. I suppose the important code will be on the PX4 middleware.

ardupilot/libraries/AP_Camera/AP_Camera.cpp

/// Servo operated camera
void
AP_Camera::servo_pic()
{
	RC_Channel_aux::set_radio(RC_Channel_aux::k_cam_trigger, _servo_on_pwm);

	// leave a message that it should be active for this many loops (assumes 50hz loops)
	_trigger_counter = constrain_int16(_trigger_duration*5,0,255);
}

ardupilot/libraries/RC_Channel/RC_Channel_aux.h

/*
  set radio_out for all channels matching the given function type
 */
void
RC_Channel_aux::set_radio(RC_Channel_aux::Aux_servo_function_t function, int16_t value)
{
    if (!function_assigned(function)) {
        return;
    }
    for (uint8_t i = 0; i < RC_AUX_MAX_CHANNELS; i++) {
        if (_aux_channels[i] && _aux_channels[i]->function.get() == function) {
			_aux_channels[i]->radio_out = constrain_int16(value,_aux_channels[i]->radio_min,_aux_channels[i]->radio_max);
            _aux_channels[i]->output();
		}
    }
}

ardupilot/libraries/RC_Channel/RC_Channel.cpp

void RC_Channel::output() const
{
    hal.rcout->write(_ch_out, radio_out);
}

ardupilot/libraries/AP_HAL/RCOutput.h

    /*
     * Output a single channel, possibly grouped with previous writes if
     * cork() has been called before.
     */
    virtual void     write(uint8_t ch, uint16_t period_us) = 0;

waterwingz wrote:
One of the kap_uav.lua users was trying unsuccessfully to engage in a dialog about the application with the pixhawk team. If you have a contact there feel free to reach out to them?

Unfortunately I don't have any contacts. I did find that Lucas De Marchi is currently working on almost all the relevant sections of the APM code, and will try to contact him.

eponymous

waterwingz
It's possible that cameras in the more expensive G series are actually calibrated - or the lens mechanism does not need calibration.

On this related note, is the G-series then the least expensive line in which you don't have to worry about this problem with "infinity" possibly not really working correctly?

waterwingz

eponymous

waterwingz
It's possible that cameras in the more expensive G series are actually calibrated - or the lens mechanism does not need calibration.

On this related note, is the G-series then the least expensive line in which you don't have to worry about this problem with "infinity" possibly not really working correctly?

I really have no idea. The G-series is the most expensive line of P&S cameras that Canon sells. So it seems likely that if any of their cameras have accurate calibration of manual focus setting it would be those ones. But I have no data (or easy way to collect any data) to support the proposition.

kona

It turns out that my S110 won't focus on infinity. The lens calibration is off. I tested two Canon S110 cameras with the exact same settings using KAP_UAV and the infinity button active. One camera takes perfect photos and the other camera produces blurry photos. Go figure...It wasn't the script. It's the darn cameras. Some are calibrated to work and some won't.

The KAP & UAV script and support is awesome! Too bad I can't say that about the gear we use it in.

Thanks again for all you guys do.

waterwingz

kona
It turns out that my S110 won't focus on infinity. The lens calibration is off. I tested two Canon S110 cameras with the exact same settings using KAP_UAV and the infinity button active. One camera takes perfect photos and the other camera produces blurry photos. Go figure...It wasn't the script. It's the darn cameras. Some are calibrated to work and some won't.

Thanks for reporting this. It's very consistent with my theory that Canon P&S manual focus settings are often poorly calibrated - even on cameras that claim to offer MF capability. And the reason you don't see this in autofocus mode is that the camera adjusts the lens position via a closed loop feedback system until it sees the sharpest image. It has no concept or need for physical distance values.

reyalp

Just FWIW, prior to this https://chdk.setepontos.com/d/12103/16 setting distances >64k on S100 through CHDK probably didn't work correctly. This would only affect CHDK overrides, it would not explain problems when setting the focus entirely through the Canon MF UI.

The Canon "focus at infinity" mode present in some Canon firmware is documented in the manuals as only being a suggestion.

waterwingz

reyalp
Just FWIW, prior to this https://chdk.setepontos.com/d/12103/16 setting distances >64k on S100 through CHDK probably didn't work correctly.

I've struggled with this. Given the depth of field of these little camera lenses, and the concept of hyperfocal distance, it seems strange that values < 64K like 50,000 or 30,000 do not appear to give sharp focus out to infinity.

« Previous Page Next Page »