I have a video of Brownian motion of microbes under a microscope. From this video, I need to calculate the angular velocity of a particular prominent cell. From my understanding, what I know is that I need to identify some points within the cell, track its location at successive time frames and thus get the velocity by dividing the distance traveled by that point by time.
Now my question is given the quality of video I have, what is the best way to track some points within the video? My quality of video is such that manual identification of points at different time snapshots within the video is not possible.
Attached a particular time snapshot of the video as an image here.
Assuming that the particular cell has only pure rotation then try following:
segment the cell in two subsequent frames ( framefrq = f).
1. Take an odd sized image matrix of the cell in the subseq frames
say c1 and c2.
2. Use the centres of the matrices as 0,0 cords.
3. Rotation matrix is R = [ cos( theta) Sin ; - sin. Cos]
4. Step theta = th1 and use
cr= Rx c1. to rotate each element of c1( i,j)
cr( ir,jr) is obtained by setting ( ir,jr) = Rx ( i,j)
so cr( ir,ij) = c1(i,j).
5. Find theta that minimises || cr - c2||
6 enjoy w= theta x f
Cheers
Another method. Assume that picel ( i,j) are not the same in all
image sequences Sk(i,j) for all k ,then with FT you can get the rotating frequencies from sequence if pixel values plotted vs k.
Cheers
I can suggest the following approach.
1. Cells detection using instance segmentation or just detection. There are a lot of approaches.
2. Track cells. The simplest approach is SORT track algorithm.
3. Analyze a cell (each track)
3.1 Align cell crops using circle detection, build "video" for each cell
3.2 Calculate optical flow
3.3 Convert the optical flow from cartesian to circular coordinates
3.4 The amplitude of result vector field is rotation speed.
Probably, you will need to crop this field or do some additional transformation in another steps.
- Badges
- Science topic
- Similar topics
- Books