I'm doing vehicles tracking project. The video is a one-way, no vehicles occlusion. After doing background subtraction, erosion and dilation, I get bounding box objects for each frame, and I can find $centroid(c_x,c_y)$ from that. So, the next thing I want to do is blobs tracking. Here is what I'm thinking on how to do the blobs tracking:

  1. At first frame, store all (cx,cy,id)
  2. Next frame, calculate distance each new (cx,cy) with each of old (cx,cy)
  3. If distance < threshold, update the old (cx,cy) which qualify the formula
  4. If distance > threshold, check if the cx,cy is near the entry coordinate (e.g. around x,y=0-640,0-40), if yes store new (cx,cy).

Is it right and possible to apply to tracking?

  • $\begingroup$ Looks good. Suggested alternative technique in answer. $\endgroup$ – Manav Kataria Jan 18 '15 at 2:02
  • $\begingroup$ What if the number of centroids increase or decrease? $\endgroup$ – Walid Ahmed Jan 5 '17 at 22:03

You might want to consider tracking using MeanShift & Cam Shift:


| improve this answer | |
  • $\begingroup$ Isn't makeshift and camshift is for single tracking, how I can use the technique for vehicles? $\endgroup$ – user3794938 Jan 18 '15 at 10:20
  • $\begingroup$ Can you elaborate on what you mean by 'single tracking'? The example in the link uses Meanshift and CamShift Techniques for vehicle tracking which I believe is your objective. $\endgroup$ – Manav Kataria Jan 18 '15 at 20:10
  • $\begingroup$ I'm sorry, I'm wrong.. I read some paper and it can tracking multiple objects.. But I'm still don't understand how to do it.. I'm confuse with the mathematics terms.. $\endgroup$ – user3794938 Jan 19 '15 at 5:09
  • $\begingroup$ It sounds good, but how do you get the bounding boxes? $\endgroup$ – user20881 May 6 '16 at 6:23

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.