People Detection in OpenCV again

There are a number of enquiries about the people detection video I did a while ago. It is a step by step explanation of what I have done. I use the XCode 4 in OSX Lion with OpenCV 2.3 to try out the following.

The first step is to download and build the latest OpenCV 2.3 into the folder at /Developer/OpenCV-2.3.0. The headers are in the include folder. Please note that you may have to copy the individual include folders from the modules folder. I build the shared libraries in the lib/Release folder.

The code is a modification of the sample peopledetect.cpp.

The second step is to display the video capture image. I use the example from the C++ reference manual in the highgui section.

#include <iostream>
#include <opencv2/opencv.hpp>
 
using namespace std;
using namespace cv;
 
int main (int argc, const char * argv[])
{
    VideoCapture cap(CV_CAP_ANY);
    if (!cap.isOpened())
        return -1;
 
    Mat img;
    namedWindow("video capture", CV_WINDOW_AUTOSIZE);
    while (true)
    {
        cap >> img;
        imshow("video capture", img);
        if (waitKey(10) >= 0)
            break;
    }
    return 0;
}

The last step is to combine the two examples into one, with a little adjustment of the detection parameters and the display rectangle size.

#include <iostream>
#include <opencv2/opencv.hpp>
 
using namespace std;
using namespace cv;
 
int main (int argc, const char * argv[])
{
    VideoCapture cap(CV_CAP_ANY);
    cap.set(CV_CAP_PROP_FRAME_WIDTH, 320);
    cap.set(CV_CAP_PROP_FRAME_HEIGHT, 240);    
    if (!cap.isOpened())
        return -1;
 
    Mat img;
    HOGDescriptor hog;
    hog.setSVMDetector(HOGDescriptor::getDefaultPeopleDetector());
 
    namedWindow("video capture", CV_WINDOW_AUTOSIZE);
    while (true)
    {
        cap >> img;
        if (!img.data)
            continue;
 
        vector<Rect> found, found_filtered;
        hog.detectMultiScale(img, found, 0, Size(8,8), Size(32,32), 1.05, 2);
 
        size_t i, j;
        for (i=0; i<found.size(); i++)
        {
            Rect r = found[i];
            for (j=0; j<found.size(); j++)
                if (j!=i && (r & found[j])==r)
                    break;
            if (j==found.size())
                found_filtered.push_back(r);
        }
        for (i=0; i<found_filtered.size(); i++)
        {
	    Rect r = found_filtered[i];
            r.x += cvRound(r.width*0.1);
	    r.width = cvRound(r.width*0.8);
	    r.y += cvRound(r.height*0.06);
	    r.height = cvRound(r.height*0.9);
	    rectangle(img, r.tl(), r.br(), cv::Scalar(0,255,0), 2);
	}
        imshow("video capture", img);
        if (waitKey(20) >= 0)
            break;
    }
    return 0;
}

Please note that the performance is pretty slow even though the capture size is 320 x 240.

Face Tracker in OSX

This video is the test run of the Face Tracker code by Jason Saragih. I compile and run it in OSX 10.7 with OpenCV 2.3.

Since Xcode will build the product into the user’s Library folder, I have to put the face model information in the product folder. In OSX 10.7, the Library folder in hidden. I have to unhide it by

chflags nohidden ~/Library/

Google Ads


Cocoa Capture and Display

The last example directly preview the camera image. In this example, I store the frame image and re-display it again using an OpenGL view. There is no QTCaptureView in the window.

The major reference is the open source Cocoa based OpenCV experiment – CVOCV.