ilchiodo's profile - activity

Hello everyone,

I made an application which detects objects of a set color in a webcam image. I implemented a simple segmentation using Mahalanobis distance (the said reference colour is "trained" from 10/20 images obtaining a mean vector and covariance matrix), and I obtained a binary image (a Mat in which pixel values are 255 or 0). How can I perform blob analysys using this binary image? Which classes/methods/functions can I use?

I'd like, for example, to determine the size of the biggest blob and ignore smaller blobs.

Thanks in advance.

I kinda solved it by myself. One of the vectors was defined as unsigned int, and I changed that to double.

Hello there! I'm trying to do an image segmentation based on Mahalanobis distance, but the implementation in Visual Studio with OpenCV 2.4.8 throws a weird exception I don't know how to solve.

Here is the snippet:

// cam_frame is a Mat object, while pixel is a std::vector made of 3 elements.

// mn is a 1x3 Mat representing the mean vector, and i_covar has been successfully computed

for (int i=0; i<cam_frame.rows; i++){   //for every pixel,
    for (int j=0; j<cam_frame.cols; j++)
    {
        pixel[0]=cam_frame.at<Vec3b>(i, j)[0];  //B component
        pixel[1]=cam_frame.at<Vec3b>(i, j)[1];  //G component
        pixel[2]=cam_frame.at<Vec3b>(i, j)[2];  //R component

        if (Mahalanobis(mn, pixel, i_cvm)<threshold)    
            mask.at<unsigned char>(i,j)=255;
        else mask.at<unsigned char>(i,j)=0;
    }
}

The Mahalanobis function throws an exception regarding the size of the arguments, but I can't really figure out.