OpenCV displays the image triplicated in a single frame - c++

Question

I am trying to build an application to simply get, save and show some frames from my camera, a DMK 41BU02 (you can consult the specifications of the device in the following link: datasheet)
My code is as simple as that:
#include "opencv2/opencv.hpp"
using namespace std;
using namespace cv;
int main(int, char**)
{
String path="~/proof.jpg";
VideoCapture cap(1); // /dev/video0 is the integrated webcam of my laptop, while /dev/video1 is the DMK41BU02 camera
cvNamedWindow( "Video", CV_WINDOW_AUTOSIZE );
if(!cap.isOpened()) // check if we succeeded
return -1;
Mat frame;
cap >> frame;
imwrite(path, frame);
imshow("Video", frame);
waitkey(0);
return 0;
}
The code compiles and executes whithout any problem, but the error arrives when the image is shown on the window or saved in the jpg file, because I get something like the following jpg, where the image is triplicated in the frame:
Resulting image of the code shown above
Some aspects to remark:
The code executes normally and returns normal images when working
with the integrated webcam of my laptop.
The DMK41BU02 camera works normally and returns normal images when working with another application, such as fswebcam or VLC.
The camera datasheet says it is compatible with OpenCV.
I have also tried the code with an infinite loop, as I know the first frame grabbed can be blank or with some type of error, but the problem is still there.
I have had some issues installing the camera drivers, but I think they're all resolved.
The laptop is a 32-bit machine with Ubuntu installed on it. Here you can see the output of uname -a: Linux AsusPC 3.11.0-18-generic #32~precise1-Ubuntu SMP Thu Feb 20 17:54:21 UTC 2014 i686 i686 i386 GNU/Linux
I have no idea of how to debug this problem and, of course, I don't know where the error could be. Could you give me any hint, please?
Thank you very much.
UPDATE: I forgot to post the weird outputs that the application writes in the terminal at the very beginning of the program:
VIDIOC_QUERYMENU: Invalid argument
VIDIOC_QUERYMENU: Invalid argument
VIDIOC_QUERYMENU: Invalid argument
VIDIOC_QUERYMENU: Invalid argument
libv4l2: error set_fmt gave us a different result then try_fmt!
HIGHGUI ERROR: libv4l unable convert to requested pixfmt
libv4l2: error set_fmt gave us a different result then try_fmt!
init done
opengl support available

Answer 1

I've had the exact same problem. The issue is within openCV itself, or more so; how cap_v4l.hpp (in the highgui module) and cap_libv4l.hpp are implemented.
The issue here is that OpenCV appearantly uses a wrong video type or channel type to read the data. Try playing arround with the different types (yuyv variants, etc) inside the opencv lib.
For some magical reason the cap_v4l.hpp is the code thats actually used by opencv and the code in cap_lib4l is not used, but seems to support more vide formats (It could be switched arround, i'm not sure about that).
Switching these files and recompiling opencv did improve stuff for me.

Answer 2

Since after the call to cap>>frame you have three channel (type=16), your capture is unaware that your camera is monochrome. Use grab-retrieve pairs instead since retrieve specifies number of channels.
bool VideoCapture::grab()
bool VideoCapture::retrieve(Mat& image, int channel=0)
Here is example code that also shows how to set camera parameters. You can also try to set some camera parameters that explicitly declare monochrome mode. If everything else fails you can always cut one image out of your triple with
Rect rect(0, 0, frame.cols/3, frame.rows);
Mat true_img = frame(rect).clone();
However, I kind of like what happens in your case. You have a natural frame queue and can analyze motion an possibly structure by looking at what happens in three consecutive frames.