Press here to go back 

Computational Neuroscience 0368.3013.01

given by Prof. H. Yeshurun

Exercise #1
Do you have more questions ? try FAQs.
a. Construct a system for detection of approaching objects, using complex log.
You may assume that the object cannot move sideways.  It may only rotate and approach or move further. (As if  an eye, or a camera is continuously  following the object).
You may do this by
    1. Performing edge detection on every image.
    2. Using complex log on the edge map.
    3. Comparing the result of every two consecutive images.

The input of your algorithm is a sequence of images of one (or more) simple object(s).
The output should be an indication of when an object approached (image number) and its location (x,y). That is, which images contained an approaching object(s) and what were the coordinates of this object.

You may write your implementation in Matlab, C++ and Pascal. For other platforms, please ask.
If you are using Matlab at home, make sure you have the 'images' toolbox, and preferably use version 5.0 or higher. This will make your life much easier.

b. Suggest a simple neural network, which implements the above task.

The images files are in this zip file:
The file also contains two matlab scripts:
CreateMovie.m may be used to create a new images series.
ReadMovie.m may be used to view the images.
The files demonstrated in class.

The exercise will be done in couples or individually.
The exercise should be printed in black and white only, and contain no more than 4 printed pages.
Please use both sides of the paper (qms11 automatically prints on both sides of every page).
You may submit the exercise until 30.5.00

If you have more questions, please contact me.