Re: [Playerstage-users] Player and OpenCv integration problem

[Jamuraa <ja...@ba...>]

On 1/5/07, chandan datta <cha...@ho...> wrote:
> Hi,
> I'm using OpenCv to do all my Image processing operations and Player for
> navigation.I'm calling OpenCv functions from Player.Here is a very basic
> client side code to make sure that I get live images from the camera through
> Player.
>
>
> #include <iostream>
> #include <cstdlib>
> #include <libplayerc++/playerc++.h>
> #include <libplayerc++/playerclient.h>
> #include <stdlib.h>
>
> #include <opencv/highgui.h>
>
> int main(int argc, char *argv[])
> {
>    PlayerCc::PlayerClient robot("localhost");
>    PlayerCc::CameraProxy cp(&robot, 0);
>
>    IplImage* img = cvCreateImage(cvSize(320, 240), IPL_DEPTH_8U, 3);
>    cvNamedWindow("camerav4l", 1);
>
>
>     uint8_t* myImage;
>    for (;;)
>    {
>       robot.Read();
>    //   cp.Print();
>
>
>       {
>          // copy Player's frame buffer to OpenCV image, converting from RGB
> to BGR
>          for (uint x = 0; x < cp.GetDepth(); x++)
>          {
>             for (uint y = 0; y < cp.GetHeight(); y++)
>             {
>                 cp.GetImage(myImage);
>                 (img->imageData + 320 * 3 * y)[x * 3]=(char)myImage;
>
>               // ((img->imageData + 320 * 3 * y))[x * 3] =  Image(320 * 3 *
> y + x * 3 + 2);
>              //  ((uchar*)(img->imageData + 320 * 3 * y))[x * 3 + 1] =
> Image(320 * 3 * y + x * 3 + 1);
>           //     ((uchar*)(img->imageData + 320 * 3 * y))[x * 3 + 2] =
> Image(320 * 3 * y + x * 3);
>             }
>          }
>          cvShowImage("camerav4l", img);
>          cvWaitKey(10);
>
>       }
>    }
>
>
>    cvDestroyWindow("camerav4l");
>    cvReleaseImage(&img);
>
>    return EXIT_SUCCESS;
> }
>
> But getting the following errors because OpenCv's image capturing
> datastructure is different from Player's.
> the error is :
>
> [root@localhost libplayerc++]# make
>
> opencv_camera.cc: In function 'int main(int, char**)':
> opencv_camera.cc:37: error: cast from 'uint8_t*' to
> 'char' loses precision
> make: *** [opencv_camera.o] Error 1
>
> Also the commented out code should work under the previous Player versions(i
> have checked the docs for presence of Image datastructure).
> So how should the datastructure of both libraris made to work together ?
>
> Also my   Machine Vision code is very CPU intensive- someone please suggest
> a coding model to be used so that vision and navigation code can be
> multithreaded and run concurrently under Player.
>
>
> Thanks for suggestions,
> Chandan.

First thing that I noticed was that you're calling the GetImage on
every iteration of that loop, which copies the image every time you
copy one pixel.  That could be a cause of your high CPU usage as well.
 Also, the myImage is not a character, it's an array of bytes.

Normally, I would write the loop as:

// copy Player's frame buffer to OpenCV image, converting from RGB to BGR
cp.GetImage(myImage);
for (uint x = 0; x < cp.GetWidth(); x++) {
  for (uint y = 0; y < cp.GetHeight(); y++) {
     (img->imageData + 320 * 3 * y)[x * 3]=(char)myImage[cp.GetWidth()
* 3 * y + x * 3 + 2];
     (img->imageData + 320 * 3 * y)[x * 3 +
1]=(char)myImage[cp.GetWidth() * 3 * y + x * 3 + 1];
     (img->imageData + 320 * 3 * y)[x * 3 +
2]=(char)myImage[cp.GetWidth() * 3 * y + x * 3];
  }
}
cvShowImage("camerav4l", img);
cvWaitKey(10);

Hope this helps!

-- 
Michael Janssen --- Jamuraa --- ja...@ba... --- ja...@de...