CS7322: Computer Vision II: Final Project Proposal


 

Predicting Trajectory of a Projectile

Michael John Cramer


 

The Idea:

For my project, I propose to implement a system that can track the trajectory of an arbitrary projectile.  Specifically, I wish to be able to predict the landing position of a projectile from a given point.  Ultimately, I would like implement a machine (a robot, perhaps) that is able to play "catch" with a user.  The user can throw a ball in the air, and the machine, predicting where the ball will land, can go there and with appropriate timing, catch the ball.

The Domain and the Scope:

This problem is definitely not a trivial matter and would require effort and material beyond the scope of this course.  To get started, however, the problem can be simplified into a much smaller domain.  First, instead of tracking an arbitrary object, I will first attempt to track something simple to recognize.  A large red ball would be ideal for this purpose.  Further, the object will be assumed to contribute no energy to its motion.  The focus can then be more directed to defining and predicting the actual motion of the object and the physics involved.  To further shrink the domain, a few assumptions will also be made.  The system will be given relevant physical dimensions at the outset, such as the height of the camera, the angle between the camera's line of vision and the ground, and the dimensions of the projectile.  This will save the system from much complex initialization and yield faster results.

I find this project an interesting approach because if I can get the system to perform to my specifiactions in real time, then I can use this advance knowledge of landing position to guide a robot to that location and attempt to intercept it.

Approach:

One goal of this project will be to implement the code on a PC.  I would like to get this system working on some hardware that a couple of my colleagues are developing for testing purposes.  Initially, I will do the coding on an SGI since the cameras and code are readily available.  As the hardware nears completion I will migrate the code to work with the PC and test when ready.

Proposed Time-line:

Here is a somewhat loosely organized plan of action:
  1. Research how to represent the object and the physics behind its motion.
  2. Begin coding and test procedures using pre-recorded demonstrations.
  3. When hardware becomes available, begin testing in real time.
Back to my class page