Download

You can download the latest snapshot just clicking HERE. Project has been finished around 20th May 2005 so there were no changes in source from then and there will probably be no more as me agreed that the project is funcional and met our needs. If you think something could be change or you changed it your self - mail us.

Below are short descriptions of three versions of the software part included in the sources package (pictures illustrating methods are at the bottom of this page). We diveded these methods into separate programs for speed purposes.

• basic algorithm [left image] - this version compare every pixel of two following vision captures. Region where the sum of differences between colours is highest is chosen for tracing. Whole analized vision is diveded into some equal rectangle regions. Grid parameters can be changed runtime, as well as difference ratio, above which we assume diff as movement (this is for vision errors minimalization - remember: me use a webcam :) )
• FRED in Monte Carlo [middle image] - this version is also based on the colours differences (R,G,B compared alone) although this time checkpoints are randomly chosen. Remember that you control your random points generator to i.e. generate more points near borders od vision. The area with highest amount of points assumed as movement is chosen for tracing
• FRED like Atari [right image] - this time before comparing anything, we create semi-vision where every small area is represented by one colour that is the average of all pixel colours in this area. Then we count differences between those areas. The zone with the highest diff is chosen.