Hi everyone. I've decided to take a brake from routines and do some cool stuff. Have you ever dreamed about future with a robots? Peaceful future where people can live alongside with robots. Smart robots. So, I would be happy to bring closer the future like this. There is a lot of things should be done to reach my goal. Lets start from something small. Lets start using Computer Vision.
OpenCV is the best open source Computer vision library. My First tutorial will be
configuring and installing OpenCV precompiled binaries with MS Visual Studio 2010.
Before we really start, let me note few more things.
This is official website of OpenCV project. There you will find a lot of valuable infos, such as:
- Documentation
- Tutorials
- OpenCV Source code
- Technical specs
- etc.
Required Software
- Visual Studio 2010
- OpenCV binaries
Second step is downloading OpenCV Library.
Once you click on requires version of OpenCV, it will redirect you on sourceforge.com, and in few seconds SaveDialog will appear on the screen.
Extract OpenCV files from archive.
Once you finished with file extracting, you need to upadate your system PATH variable. In my case it's "C:\OpenCV\opencv\build\x86\vc10\bin"
OK, first part of the job already done.
Now we should open Visual Studio, create new C++ project, and configure it to use OpenCV.
- Run Visual Studio
- Create new C++ project, select "Win32 Console application"
- Set Project name (I've used "Tracker2")
Make sure you are select "Console application" type. Project should be empty.
Great, Project created. Now we need C++ code file for our "OpenCV Hello World " application.
So, Lets adding new C++ file into the project. (I called it same as Project-Tracker2)
Now we need to configure our project to work with OpenCV libraries.
- Go to "Property Manager Window"
- Select Debug folder.
- Right click and select "Add new project property Sheet".
- Go to VC++ Directories" - "Include Directories"
Add path "c:\OpenCV\opencv\build\include\"
- Go to "Library Directory"
- Add "C:\OpenCV\opencv\build\x86\vc10\" path to OpenCV libraries
- Go to "Linker- Input - Additional Dependencies"
- Add "*d.lib" files from "C:\OpenCV\opencv\build\x86\vc10\lib\ folder"
- *d- means- Debug version.
- So in my case, I've added next set of files:
- opencv_calib3d246.lib
- opencv_contrib246.lib
- opencv_core246.lib
- opencv_features2d246.lib
- opencv_flann246.lib
- opencv_gpu246.lib
- opencv_haartraining_engine.lib
- opencv_highgui246.lib
- opencv_imgproc246.lib
- opencv_legacy246.lib
- opencv_ml246.lib
- opencv_nonfree246.lib
- opencv_objdetect246.lib
- opencv_ocl246.lib
- opencv_photo246.lib
- opencv_stitching246.lib
- opencv_superres246.lib
- opencv_ts246.lib
- opencv_video246.lib
- opencv_videostab246.lib
That's all, my congratulations. I did it.
Just for testing purpose, we will create small "Hello world" app using our OpenCV configuration.
We will try to identify round object with solid colour. Not a best solution, but good enough for first time.
Open Tracker1.cpp file and paste next code:
// Tracker7.cpp
#include<opencv/cvaux.h>
#include<opencv/highgui.h>
#include<opencv/cxcore.h>
#include<stdio.h>
#include<stdlib.h>
///////////////////////////////////////////////////////////////////////////////////////////////////
int main(int argc, char* argv[]) {
CvSize size640x480 = cvSize(640, 480); // use a 640 x 480 size for all windows, also make sure your webcam is set to 640x480 !!
CvCapture* p_capWebcam; // we will assign our web cam video stream to this later . . .
IplImage* p_imgOriginal; // pointer to an image structure, this will be the input image from webcam
IplImage* p_imgProcessed; // pointer to an image structure, this will be the processed image
/* IPL is short for Intel Image Processing Library, this is the structure used in OpenCV 1.x to work with images */
CvMemStorage* p_strStorage; // necessary storage variable to pass into cvHoughCircles()
CvSeq* p_seqCircles; // pointer to an OpenCV sequence, will be returned by cvHough Circles() and will contain all circles
// call cvGetSeqElem(p_seqCircles, i) will return a 3 element array of the ith circle (see next variable)
float* p_fltXYRadius; // pointer to a 3 element array of floats
// [0] => x position of detected object
// [1] => y position of detected object
// [2] => radius of detected object
int i; // loop counter
char charCheckForEscKey; // char for checking key press (Esc exits program)
p_capWebcam = cvCaptureFromCAM(0); // 0 => use 1st webcam, may have to change to a different number if you have multiple cameras
if(p_capWebcam == NULL) { // if capture was not successful . . .
printf("error: capture is NULL \n"); // error message to standard out . . .
getchar(); // getchar() to pause for user see message . . .
return(-1); // exit program
}
// declare 2 windows
cvNamedWindow("Original", CV_WINDOW_AUTOSIZE); // original image from webcam
cvNamedWindow("Processed", CV_WINDOW_AUTOSIZE); // the processed image we will use for detecting circles
p_imgProcessed = cvCreateImage(size640x480, // 640 x 480 pixels (CvSize struct from earlier)
IPL_DEPTH_8U, // 8-bit color depth
1); // 1 channel (grayscale), if this was a color image, use 3
cvWaitKey(2000);
while(1) { // for each frame . . .
p_imgOriginal = cvQueryFrame(p_capWebcam); // get frame from webcam
if(p_imgOriginal == NULL) { // if frame was not captured successfully . . .
printf("error: frame is NULL \n"); // error message to std out
getchar();
break;
}
//cvInRangeS(p_imgOriginal, // function input
// CV_RGB(175, 0, 0), // min filtering value (if color is greater than or equal to this)
// CV_RGB(256,100,100), // max filtering value (if color is less than this)
// p_imgProcessed); // function output
//cvInRangeS(p_imgOriginal, // function input
// CV_RGB(100, 10, 20), // min filtering value (if color is greater than or equal to this)
// CV_RGB(255,50,50), // max filtering value (if color is less than this)
// p_imgProcessed); // function output
cvInRangeS(p_imgOriginal, // function input
CV_RGB(110, 0, 0), // min filtering value (if color is greater than or equal to this)
CV_RGB(255,50,50), // max filtering value (if color is less than this)
p_imgProcessed); // function output
p_strStorage = cvCreateMemStorage(0); // allocate necessary memory storage variable to pass into cvHoughCircles()
// smooth the processed image, this will make it easier for the next function to pick out the circles
cvSmooth(p_imgProcessed, // function input
p_imgProcessed, // function output
CV_GAUSSIAN, // use Gaussian filter (average nearby pixels, with closest pixels weighted more)
9, // smoothing filter window width
9); // smoothing filter window height
// fill sequential structure with all circles in processed image
p_seqCircles = cvHoughCircles(p_imgProcessed, // input image, nothe that this has to be grayscale (no color)
p_strStorage, // provide function with memory storage, makes function return a pointer to a CvSeq
CV_HOUGH_GRADIENT, // two-pass algorithm for detecting circles, this is the only choice available
2, // size of image / 2 = "accumulator resolution", i.e. accum = res = size of image / 2
p_imgProcessed->height / 4, // min distance in pixels between the centers of the detected circles
100, // high threshold of Canny edge detector, called by cvHoughCircles
80, // low threshold of Canny edge detector, called by cvHoughCircles
4, // min circle radius, in pixels
400); // max circle radius, in pixels
for(i=0; i < p_seqCircles->total; i++) { // for each element in sequential circles structure (i.e. for each object detected)
p_fltXYRadius = (float*)cvGetSeqElem(p_seqCircles, i); // from the sequential structure, read the ith value into a pointer to a float
printf("ball position x = %f, y = %f, r = %f \n", p_fltXYRadius[0], // x position of center point of circle
p_fltXYRadius[1], // y position of center point of circle
p_fltXYRadius[2]); // radius of circle
// draw a small green circle at center of detected object
cvCircle(p_imgOriginal, // draw on the original image
cvPoint(cvRound(p_fltXYRadius[0]), cvRound(p_fltXYRadius[1])), // center point of circle
3, // 3 pixel radius of circle
CV_RGB(0,255,0), // draw pure green
CV_FILLED); // thickness, fill in the circle
// draw a red circle around the detected object
cvCircle(p_imgOriginal, // draw on the original image
cvPoint(cvRound(p_fltXYRadius[0]), cvRound(p_fltXYRadius[1])), // center point of circle
cvRound(p_fltXYRadius[2]), // radius of circle in pixels
CV_RGB(255,0,0), // draw pure red
3); // thickness of circle in pixels
} // end for
cvShowImage("Original", p_imgOriginal); // original image with detectec ball overlay
cvShowImage("Processed", p_imgProcessed); // image after processing
cvReleaseMemStorage(&p_strStorage); // deallocate necessary storage variable to pass into cvHoughCircles
charCheckForEscKey = cvWaitKey(10); // delay (in ms), and get key press, if any
if(charCheckForEscKey == 27) break; // if Esc key (ASCII 27) was pressed, jump out of while loop
} // end while
cvReleaseCapture(&p_capWebcam); // release memory as applicable
cvDestroyWindow("Original");
cvDestroyWindow("Processed");
return(0);
}
Build and run app.
As result you should get something like this:
Feel free ask any questions.
