|
Multiple camera problems
Jim from United Kingdom [12 posts] |
14 year
|
|
I've got 2 identical webcams, that use the Microsoft camera driver on Vista 64. They both work individually in Roborealm (i can switch between them in Options - 1 is listed as "USB Camera" the other as "USB Camera #2") however if i try to mosaic them, or use markers to switch between them on the fly only one will work - it's not always the same one, it depends which is active first. The one that doesn't work is blank, and frame rate on the working camera is very poor, also the led on the webcam that's not working flashes (no mention in the manual as to what that might mean...). I'm guessing this is the cameras, as other people don't seem to be having issues? I was using 2 different cameras with success before, but the difference in images between the 2 made it time consuming to work with.
|
|
|
|
Jim from United Kingdom [12 posts] |
14 year
|
|
I've just tried with the cameras in separate hubs, and location-switching is working now - maybe some sort of usb address issue...
|
|
|
|
Anonymous |
14 year
|
Jim,
Thanks for the follow-up. There are some cameras that don't work well with more than one of the same brand but I'm not sure if anyone has fixed this by using a separate hub. Unfortunately while the USB spec does include a serial id number in each device this is not required nor guaranteed to be unique. So most likely the USB hub creates that uniqueness since it itself also has an id. Strange that it works but thanks for the tip. That will be useful for others!
STeven.
|
|
|
|
Jim from United Kingdom [12 posts] |
14 year
|
Cheers. I've got a couple of other questions - firstly i've had a couple of crashes - usually doing silly things, is knowing when these happen useful to you? (i tried entering global variable names typed directly into X and Y boxes for an edge detector which crashed to desktop - the fact that it had up/down arrows should've been a clue that it wanted a number - the crash is repeatable here) On that note i'm trying to find the horizontal width of found blobs - is scanning for edges along the line of it's Y midpoint a reasonable way to go about it? I've got stereoscopic vision for distance sensing working to within few mm, and i'd like to be able to calc object size from it.
Also, i use a screen quite a way away, and the text in the vbscript interface is almost unreadable - some controls for the textbox would be lovely (as would syntax highlighting and the ability to tab, but i can always use notepad++ for bigger bits of code) - just an updown control for text size in box is all i need.
Ttbh i'll be buying it once my 30 days are up regardless, it's a great bit of software, so thanks :)
|
|
|
|
Anonymous |
14 year
|
Jim,
By all means let us know about any crashes that you find. If we can reproduce them we normally can fix them right away. We naturally test for many situations but sometimes things do slip by.
What edge detector module are you using to generate this crash?
You probably want to use the Geometric_Statistics module to calculate the width and height for each blob ... assuming a blob's boxed dimensions are what you want. If you are looking for the widest point in the blob let us know and we can easily add that as part of the geo stats.
We will look into adding a font size, allow for tabs and syntax highlighting. But keep in mind that the provided text area is just for simple editing as you have noted that you can always use a more full featured editor for longer scripts.
STeven.
|
|
|
|
Jim from United Kingdom [12 posts] |
14 year
|
The crash is using the Edge Probe tool - i wanted it to scan a horizontal path based on variables returned by a VBScript block, so tried just typing the variable name into the X and Y boxes.
I've also got a color gif that crashes the Fiducal training - it works fine if i just save it as a jpg (gif attached if you're interested). Again i'm probably going about things wrong but it consistently crashed to desktop with it as a gif.
I've found the geometric statistics, that's near enough what i need, i can extrapolate/calculate anything else i want from that.
It did occur to me after posting that any time you spend on the scripting box could potentially be better spent on Vision-related stuff, and i can just use notepad++ and stop being lazy, so on that basis i withdraw the request for a better scripting box...
Cheers for all the help Steven :)
|
|
|
|
Jim from United Kingdom [12 posts] |
14 year
|
|
Just read the fiducials tutorial properly, and seen where i'm going wrong - the crash is odd but doesn't really matter, my image would never work like that anyway. I've printed some of the test fiducials and i'm more than happy with the results :)
|
|
|
|
Anonymous |
14 year
|
Jim,
We got the fiducial error fixed. That was due to the file being somewhat small.
We also updated the scripting box to include a font size selector, a tab size selector (tabs now work in the box) and also a line number indicator. Yes, while these things do turn our attention away from other modules these additions seemed like good ideas and were also on our todo list (i.e. we also were annoyed by the lack of tabbing and line numbers). We'll wait on the syntax highlighting as that's a bit more involved.
Once you download the new version can you try the crash using the edge probe? We can't seem to get that to crash. If it still does can you save the configuration to a robofile that you use just before typing in the variable name that causes the crash? In that way we should be able to replicate the error.
v2.18.6 has these changes.
Thanks,
STeven.
|
|
|
|
Jim from United Kingdom [12 posts] |
14 year
|
The script box changes are great, thanks :)
The edge probe now only crashes with a few variables that I renamed to start with a + symbol to float them up to the top of the variable list - i can't remember if that was the case before (i don't think so, but it would make sense...) - I've renamed them from "+Right X" to "_Right X" so they still come to the top, and they also work as inputs for the edge probe. So again, absolutely great, thanks.
On the other hand, the fiducial module's not writing the text labels to the screen for me any more - it's boxed in a couple of the markers, but the "FIDUCIAL CONFIDENCE" variable is -1, all the others empty. In fact, now i look at it although it trains and lists the right number of fiducials, it lists the number loaded as 0 in the main fiducial screen after. I've cleared the folder right out (including the roborealm.fiducials file) and left just one of the original images in there, and re-trained but still no joy.
Cheers,
Jim
|
|
|
|
Anonymous |
14 year
|
Jim,
Opps, our bad. In correcting the previous error we were cropping the image incorrectly before comparison. Can you redownload and try again?
Yes, it is possible that variables named as "+Right X" would probably not work correctly. We tried a couple combinations to get RR to crash but no luck. In general I would avoid using any mathematical symbols in variable names (like +-*/, etc) and stick with letters, numbers and the underscore. I'd also avoid using spaces in variable names. To be completely compatible with other languages and platforms I'd also not start the variable name with a number. In this way you can be sure to use the same variable name in many other platforms other than just RoboRealm. Which is just a good practice in general.
Let us know if anything else comes up.
STeven.
|
|
|
|
Jim from United Kingdom [12 posts] |
14 year
|
It works now once after i install, but if i re-train it, it goes back to "Loaded fiducials: 0". I've just got the default fiducials in the folder now.
Cheers,
Jim
|
|
|
|
Anonymous |
14 year
|
Jim,
Does the scroll bar show up when training and then somehow just not report any of the images found? Are there any images in the folder that is being trained on?
STeven.
|
|
|
|
Anonymous |
14 year
|
The trainer reports the right number of images (just using the default images, in the folder they are installed to) - 18 processed, 0 remaining, but clicking OK after training returns to the Fiducial module screen, with "Loaded Images" listed as 0.
Re-installing gets the default markers workng again until i re-train - the roborealm.fiducials generated when i train is also much smaller than the one the installer leaves there.
Cheers,
Jim
|
|
|
|
Jim from United Kingdom [12 posts] |
14 year
|
I've gone back to an older version for now for the fiducials and it's working ok for all the default images, but the fiducial array i get back seems to go
confidence,
X trans,
y trans,
scale,
X rot,
y rot,
orientation
path start inedx
path length
X1
Y1
X2
Y2
X3
Y3
X4
Y4
Which is slightly different from the documentation, also if there's only one fiducial found it ommits a path length, which pushes the rest out of order.
Cheers,
Jim
|
|
|
|
Jim from United Kingdom [12 posts] |
14 year
|
|
I'm getting them in the correct order now - both times i was just using the Write Variable to output the whole FIDUCIALS array to a text file so i could seee what the format was - it now matches the documentation so not sure what was happening there - i also get the correct indexes now. I was testing with just one image in the folder, which might have caused it?
|
|
This Thread