Control & Robotics Lab Presented By: Yishai Eilat & Arnon Sattinger Instructor: Shie Mannor Project Presentation
System Setup camera
Objectives Locating a ball in a Foosball table based on a video stream Real time performances A robust solution Simplicity
The Solution Tracking & Estimation process Increase success probability Enable limited search Searching the ball in a restricted area Reduce calculation Time Eliminate irrelevant areas
Tracking & Estimation sequence Based on continuity Linear movement Needs history v
Search in full size window Calc. Movement Vector Search in window around the estimated position Found? Enlarge Window No Update & Go to Next Frame Yes The Main Loop
Problems in Finding The Ball Smeared ball Eclipsed ball Black & white picture Noises Real-Time
The Main Idea Find Pixels Above Threshold = Candidates Filtering: Form Objects Subtract a const Background Noise Players Decide Who is the ball
Players Filter Identify pattern of players. Based upon location Assumes a symmetric Table Doesn ’ t Filter The Keepers
Decision part Rule out: objects that are too small objects in keeper zone (if an object outside the Keeper zone exists) Chose the closest object to the Estimated Position
Live show The short clip will demonstrate the various features we discussed.
Future Improvements The Table The Camera Software optimization Integrate mechanic sensors
Thank you ! The End.