Introduction to Blob Computation and Tracking

Up to now we have already compute a color region of interest. We are ready to start using a sequence of images with the main objective of tracking some moving object. At each frame we have to determine where the object is moving or -in the case of a mounted camera - which is the motion we have done.

Mission

In this Lab we like you to write an efficient program to track a region in a sequence of images. Make use of the a convex hull to define the region envelope and by means of conecting lines create an envelope of such a region. Your algorithm needs to deal with some deformations of such an initial Blob.

• What happens if such a blob becomes partially occluded ?

Starter Code

The Convex Hull

convex_hull (int *rows, int *columns, int n)

{

   int i,j,k;
   double angle_2pt(), prev, best, x;
   k = 0;
   for (i=1; i<n; i++)
   if (rows[i] > rows[k]) k = i;
   else if ((rows[i] == rows[k]) && (columns[i]<columns[k]))
       k = i;			/* Same row, choose leftmost */
       hswap (rows, columns, k, 0);
       rows[n] = rows[0]; columns[n] = columns[0];
      prev = -1.0;	j = 0;
 do {
    best = 360.0;  k = -1;
    for (i=j+1; i<=n; i++) {
        x = angle_2pt (rows[j],columns[j], rows[i],columns[i]);
        if ( (x>prev) && (x<best) ) {
            k = i; best = x;
        } 
        else if ( (x>prev) && (x == best) ) {
       if ( (abs(rows[i]-rows[j])+abs(columns[i]-columns[j])) >
         (abs(rows[k]-rows[j])+abs(columns[k]-columns[j]))) {
       k = i; best = x;
       }
    }
  }
  if (k > 0) {
  prev = best;
  j = j + 1;
  hswap (rows, columns, k, j);
}

}

while (k>0 && (j<n));
rows[j+1] = rows[0]; columns[j+1] =columns[0];
return j+1;

}

hswap (int *rows, int *columns,int i,int j) {

int t;
t = rows[i]; rows[i] = rows[j]; rows[j] = t;
t = columns[i]; columns[i] = columns[j]; columns[j] = t;

}

Draw a Line

The Kalman Filter

Here you will find some codes for the implementation of the Kalman Filter

Speed Estimation from Single Loop Data

Modelling Camera Model

Links & References

Hints