dmngu9's profile - activity

Hi, i have implemented particle filter for tracking an object. It works well when the object is stationary , at that instance the particle converge . However, when i move the object to another location suddenly, the particles cant move .

This is because i eliminate low weighting particle and replace them with high weighting ones at first object's position which results in particle concentrating in that position. When i move the object to position that is far away from the first position, the particle swarm is not spread out enough to converge to the second position.

Any suggestion to improve on the resampling part, i use stratified resampling . help me please on the redistribution of particles

 Mat image; 
  Mat grayFrame;
  Mat blue_matrix ;
  Mat diffFrame;
  image = cv_ptr->image;
  float max=0;
//obtain luminosity image and blue matrix and diffframe of blue and luminance 
  for (int row = 0; row<=479; row++)
  {
    for (int col =0; col<=639; col++)
    {
        Vec3f color_value = image.at<Vec3f>(row, col);
        int blue = color_value.val[0];
        int green = color_value.val[1];
        int red = color_value.val[2];
        grayFrame.at<int>(row,col) = round (0.21*red + 0.72*green + 0.07 * blue); 
        blue_matrix.at<int>(row,col) = blue;
        cout << blue<< "       "<<endl;
}}

Hi, I have a bgr image and want to split into r g b matrice and convert original matrix to grayscale. However , this code gives me core dump error because of these lines

grayFrame.at<int>(row,col) = round (0.21*red + 0.72*green + 0.07 * blue); 
        blue_matrix.at<int>(row,col) = blue;

Also, i notice the blue value is weird. This is the value i obtained -2147483648 -2147483648.

Anyone help me with this problem please

Hi i tried the tutorial opencv surf but i get error descriptor empty. I want to skip the frame which gives empty descriptor. after trying a few approaches, it still keep terminating the process without jump to next frame when that error occur. any suggestions

also, i have this error as well OpenCV Error: Assertion failed (count >= 4) in cvFindHomography . I havent found the reason for this error

thanks for your help. but my question is that 2 frames rgb and depth both have same resolution. if i have point (u,v) in rgb frame, is the point (u,v) in registered depth map corresponding to the point(u,v) in rgb frame (in other words aligning with each other ).

I have RGB and depth frame both have resolution of 480x640 pixels. I already align these 2 frames together and what i obtained is the pointcloud rgb-d . How can i implement SURF with these RGB-D image since SURF opencv only for 2d image?

I am also confused about this situation. If i have a point in RGB image, lets say (100,200). Now i want to find the depth of that point. Can i just go to the point (100,200) in depth registered image and get the depth out?

I use ROS Kinect OpenNI

Please let me know

can i ask you about point2f? When i declare cvPoint(200,100), it means the point is at rows 200th and column 100th right? But sometimes people declare as double i =100.09, double j = 300.5 and cvPoint(i,j)? What does Point2f means?

I thought the coordinates of a point is based on rows and columns of the image. And decimal values should not exist

im using rgb image? sometimes it works, sometimes it doesnt

This might be a dumb question. In the tutorial, if i have 2 same pattern but with different colors, Does that affect the keypoint and matching in SURF?