I tried “cam_setLightMode 2” and didn’t see much difference. It did give me a response 5: 0 (0x0). There may have been slightly less washout, but outdoors in bright sunlight mostly everything is washed out. If it’s overcast I get some detail, but even then things that are not completely in shadow are washed out.
Thanks for trying-- we’ll look into this.
Any updates on this issue? I really need to find a way to get usable tracking in outdoor lighting conditions.
Wayne
Popping stuff off the stack— gonna try to get to it this week.
Wayne Holder wrote:
Any updates on this issue? I really need to find a way to get usable tracking in outdoor lighting conditions.
Wayne
It’s been over a week. Any results on this, yet?
The latest firmware seems to help some (firmware-0.1.44.hex), but doesn’t completely solve the problem.
Dude! Don’t forget about the “Brightness” parameter. Try decreasing it if it’s too bright!
I have found good results since the FW update by setting the cam_setBrightness to 50 or lower, setting cam_setMode to 1, and
setting cam_LightMode to 2. (Even in a darker or subdued lighting room), for some reason it seems to make the Auto Brightness
respond faster.
It also seems to make any “hue rich” object (blues, reds, greens) stand out better for signature capture. (At least it does for me)
Regards,
Dan
Cool. But I suspect that the cam_setMode, cam_setLightMode are not having much effect. These are updated automatically now. Setting the “Brightness” parameter in the configure dialog saves the brightness value to flash (and restores the next time you power up), whereas setting cam_setBrightness from the command prompt will only set the current value (and not save to flash).
Here is a sample with brightness set to 50. (See example Nowashout.jpg)
The white is SNOW, and there is complete, direct sunlight shining down on it. The image is not washing out.
I could have gone to 60 or possibly higher with the brightness.
There is an unfortunate side effect though. I took a bright green object outside to test signature recognition,
and in the sunlight (without washout), the camera had an almost impossible time capturing the Color Saturation.
When it finally did capture a signature, it interpreted everything that was not black (or very dark) to being the green hue.
This time of year up where I am, the only thing outside that was bright green, was the ball I was holding…
I also tried adjusting the Min saturation up to 50 in increments of 5, no help.
I took the saturation spread down to .005 in small increments as well to no avail.
Regards,
Dan Thimm
The white snow reflecting the sun causes the rest of the image (the parts that aren’t white snow reflecting the sun) to be underexposed (because the snow is so bright). It’s probably the most challenging scenario imaginable. Photographers hate this situation too because it’s too “contrasty”.
A possible way to deal with this is to crank up the brightness so that the underexposed parts are less underexposed. Training on objects that appear dark in PixyMon is not going to work.
Just to be clear, if I can adjust the brightness using Pixymon, will it say set to the last brightness setting when I plug my Pixy into an Arduino?
Wayne
That’s correct— all modifications you make to parameters in the Configure dialog (e.g. “Brightness”) will be saved to flash on the Pixy. (and used the next time your power up Pixy)
Thanks for the confirmation on having the brightness setting saved to flash. Is there any prospect for a more dynamic method of brightness control? For example, using a histogram analysis, or something similar to set exposure. Ideally, I’d like to have the Pixy be able to automatically adapt to different lighting conditions. Or, lacking that, is there anyway an Arduino program could adjust the exposure, perhaps using an ambient light sensor, or light meter to gauge lighting conditions and then adjust the brightness setting in the Pixy.
Wayne
Hi Wayne,
That’s a good question. Changes in the brightness of a given object in the image change based on 2 things— 1) the ambient light changing or 2) the exposure of the camera changing-- ie taking into account a change in the average brightness caused by, for example, a window in the image that wasn’t there before or a darker object appearing in the image, etc. We can’t do much about (1), but for (2) A smarter thing would be to adjust the exposure based on the regions of interest. We have some good ideas on this, and hope to have something released in the coming months.
We can also add the ability for the Arduino to adjust the aggregate brightness. But it then becomes a problem of when to adjust-- how will your Adruino program know? We trade one problem for another.
Can you describe your application? Are you moving between two different lighting conditions? (e.g. outdoor and indoor?)
Can you describe your application? Are you moving between two different lighting conditions? (e.g. outdoor and indoor?)
My application is a mobile robot where I want to use the Pixy for obstacle and marker detection. My robot will be typically operating outdoors in blue sky conditions. The main concern I have is dynamic shading from cloud movement. My thought was that it might be fairly easy to a light meter chip, such as the BH1750, and use the value it computes to adjust the Pixy.
In outdoor conditions, certain portions of the image are always going to “blow out” due to overexposure. The problem is that these areas tend to cause the camera to darken the image way too much in compensation. So, ideally, if the Pixy could be set to ignore these blown out regions, the resulting brightness setting might allow better hue tracking in the portions of the image that remain visible.
Wayne
Hi to all,
as I can see, it is several months after your last post. I am also very interested in outdoor robotics application of Pixy.
Could you share some results of your testing of Pixy outdoors? For instance video of outdoor performance would be very helpful.
Also: how is it possible to use the Pixy for obstacle detection??
Thanks in advance!
Jarda
Using the latest firmware (0.1.44), I’m getting a very washed out image (see attached). The trees are gray and the sky is a gray-blue. Is there something I can do to correct this?
I’ve played with all of the controls I could find: cam_setBrightness, cam_setMode, cam_setLightMode, and it doesn’t get any better than this.
Thanks,
Ag Primatic
This may be a solution that would work for your outdoor application.
bit.ly/IR-LOCK
Best,
Thomas
Ag Primatic,
your picture looks bizarre. A couple questions:
Are you using the lens that came with Pixy? (it looks like your lens isn’t blocking IR light)
Do you see the same effect indoors? (can you include a picture?)