Hi
I'm working on a pandaoard to make pattern recognition.
I have install ubuntu 11.04 and OpenCV 2.3 on it.
when i run programs on pictures, all goes right but when I want to use
vidéos ther is this warning message:
no accelerated colorspace conversion found from yuv420p to rgb24
and for a simple programme like resize the vidéo, it run slowly.
I googled this message and I have found that it can be solved by
updating the ffmepg library.
So I've done it thanks to this web site :
http://ubuntuforums.org/showthread.php?t=786095
After that, I had to reinstall OpenCV and the compilasion stop at 36%
with an error that told me that ther is a problem with the ffmpeg
library.
So google my friend told me after some research that it's a problem
with a file in opencv/modules/highgui/src/cap_ffmpeg.cpp. So I have
modified cap_ffmpeg.cpp like it says here :
https://code.ros.org/trac/opencv/attachment/ticket/1020/ffmpeg_build.patch
But after that, I had a new problem!!!
/usr/bin/ld: /usr/local/lib/libvacode.a(libvacode.o): relocation
R_X86_64_32 against `a local symbol' can not be used when making a
shared object; recompile with -fPIC
/usr/local/lib/libvacode.a: could not read symbols: Bad value
make[1]: *** [libavutil.so.49] Error 1
make: *** [lib] Error 2
So I have googled again and I found that this problem can be solved in
changing the config of ffmpeg with the -fPIC option.
After that, all goes well for the compilation of openCv but when I
have compiled again my programme ther still was the same warning
message :
no accelerated colorspace conversion found from yuv420p to rgb24
all this... for nothing!!
So if someone have a solution to solved this problem I will be very
happy!!
I'm open for all the solutions to run my programme on the pandaboard
even if I have to change OS.
thanks in advance for your help
pierre