Vision Tracking System Presented By Timothy Bagnull James Deloge Chad Helm Matthew Sked ECSE 4962 – Control Systems Design Rensselaer Polytechnic Institute 4/22/03 Introduce team etc.
Overview Objective and Specifications System Design Testing and Verification Problems Encountered System Demonstration Conclusions State topics to be covered
System CAD Model
Objective and Specifications Track a moving point with a camera and pan-tilt system. Controller Specifications Maximum target speed: 1 ft/s Settling time: 0.1 s Overshoot: 2 % Vision Specifications Initialize system using an edge detection algorithm Track target using a Kalman Filter Re-cap design objective
System Design: Controller Linear Controller Effects of Coulomb Friction Real Time System Response Motor Saturation
System Design: Controller Linear Controller - Pan
System Design: Controller Linear Controller - Tilt
Effects of Coulomb Friction
Effects of Coulomb Friction
System Design: Controller Real Time System Response - Pan
System Design: Controller Real Time System Response - Tilt
System Design: Controller Motor Saturation - Pan
System Design: Controller Motor Saturation - Tilt
System Design: Vision Implemented using C++ 4 levels of communication Camera – Frame Grabber – Computer - ARCS Find the target: Roberts Edge Detector Track the target: Incremental Step Function Future Modification: Kalman Filter, Pattern Recognition
System Design: Vision Roberts Edge Detector Calculates the first order image gradient magnitude Through a threshold function we determine which pixels are line pixels and which are not By assuming an ideal environment we can calculate the center of the point by taking the mean of our line pixels
System Design: Vision Normal Lighting Conditions Original Screen Grab | Edge Detection Output
System Design: Vision Poor Lighting Conditions Original Screen Grab | Edge Detection Output
Original Screen Grab | Edge Detection Output System Design: Vision Focus Conditions Original Screen Grab | Edge Detection Output
System Design:Vision -,- +,- +,+ -,+ Incremental Step Function Determines target position in coordinate frame Steps towards target using increment function (0,0) (640,0) -,- +,- (320,240) +,+ -,+ (480,0) (640,480)
Testing and Verification Controller Trajectory program Line,Circle, Jog Functions Vision Edge Detection Incremental Step Function System
Problems Encountered Real time system controller tuning vs. simulated controller tuning Coordinate transformations between vision and ARCS systems Learning programming interfaces between mechanical and visual systems
Final System Performance Final Performance Maximum tracking speed: 0.5 ft/s Settling time: 1 s Overshoot: 50% Initial Specifications Maximum target speed: 1 ft/s Settling time: 0.1 s Overshoot: 2 %
Open Loop Response Pan: Torque at 0.1 Nm
Open Loop Response Tilt: Torque at 0.09Nm
Conclusion We successfully implemented a vision system with a mechanical pan/tilt Future work can be done to make this system much more robust Overall we have shown that vision can be a used as an effective sensor in controls
Demonstration Independent joint test System test Track horizontally moving target Track vertically moving target System test Random Motion Tracking performance test
Questions?