Multiple objects tracking

Here, I just want to say some ideas about this research point.

In my opinion, what is tracking? Just want to find some one and to tracking him and to find what he do or where he go.

This video often is low correlation camera videos. If we manually to find this people, this can be done well.

However, this work will cost wait time by people. Human is so laze that we want to do this work by compute.

So this research point is researched by other people.

However, please people must know some challenges about this research such as the occlusion.

How to handle the occlusion problem?

Please people proposed into methods to handle this problem. Such as some people approach the people model to detection the occlusion

Person in the video frame. And other people utilize some good feature point or pixel to do match between different frame to find the same object.

In the sight, some people utilize part detection method. Some researchers apply the part matching methods.

However, I think we do not find a good method.

Because, if human do not find one person by our eye, I think compute must find this person too.

What to match?

The tracking problem is one matching problem. We need to find one robust feature to represent one person so that this person can

Represent one person so long time until this person disappearance. So people begin different feature to represent one person such HSV.

HSV. Hog ans so on. However, for the nonrigid person body, it is so hard to find the robust feature. In order to handle this problem,

Some researchers utilize the online training to handle this problem such as TLD method. However, I find that this method only get a good

Performance in some special video data. So we also do not know how to handle this problem.

In the produces, some people begin to combine different methods to handle the tracking problem. Such as feature points + particle filter.

SVM + detection and so on.

Until now. I find that this is so hard to handle this problem. I want to give up this research. I find that I do not find hope in this research.