For this problem set your goal will be to implement a simple person
tracker. You may use the language and environment of your choice to
solve the sub-problems, although popular candidates would be OpenCV or
MATLAB. You are free to use libraries of existing computer-vision
algorithms.

Problem 2.1)

Download the following uncompressed AVI to use as training data:

http://www.k2.t.u-tokyo.ac.jp/members/carson/rs11/20110712-ps2.avi

Verify that you can open and play the video on your computer. For each
frame of the video, find the Harris corners. Plot the number of
corners found over time.

Problem 2.2)

Compute the optical flow of the video sequence. Estimate the velocity
of the person in pixels per second. If an object of known size
were in the image could you estimate how many meters per second the
person is moving?

Problem 2.3)

Using a technique of your choice identify the background of the
training data video. Show a still of the 60th frame of the video with
the background pixels subtracted.

Problem 2.4)

Draw a bounding box around the area which contains the person in each
frame. Assuming that the camera is stationary, does your approach work
on other video sequences?