Download presentation
Presentation is loading. Please wait.
1
1cs426-winter-2008 Notes Assignment 1 is out, due immediately after reading week (Feb 25) Please read: William T. Reeves, "Particle systems: a technique for modelling a class of fuzzy objects", SIGGRAPH 1983 Gavin Miller and Andrew Pearce, "Globular dynamics: a connected particle system for animating viscous fluids", SIGGRAPH 1989
2
2cs426-winter-2008 Match move Matching up CG camera to real-footage Easy version: Have rough CG model corresponding to real scene already, just solving for camera Hard version: Need to figure out CG model as well
3
3cs426-winter-2008 Match points Need to identify image space positions of enough world space points in each frame Technically only need 3 non-collinear if field- of-view is known, 4 if not More points are essential for robustness Also deal with camera distortions Typically identify points by hand For difficult scenes (grass?) may need computer vision techniques, or just put stuff in the scene to track (and paint over later)
4
4cs426-winter-2008 Solving match move This essentially boils down to the same problem as IK, though numerically harder Ask the computer to find camera parameters which give as good a fit for all the match points as possible Since some match points may be misidentified, need to robustly deal with outliers (discard suspicious data) May need interactive help from user to lock on (first guess at camera parameters etc.) It helps a lot if real-life measurements made to determine exactly where match points are in world-space But if not, can still reconstruct it from the data…
5
5cs426-winter-2008 3D reconstruction Leads into another hot topic: 3D model reconstruction Active scanning: stick object (or person) in 3D scanner Can get very high accuracy Passive reconstruction: try to build a good CG model from footage (or extra photos) itself Much harder, much more useful, slow progress being made
6
6cs426-winter-2008 Procedural Animation
7
7cs426-winter-2008 Particle Systems For fuzzily defined phenomena, highly complex motion, etc. particle systems provide a (semi-)automatic means of control Break up complex phenomena into many (hundreds, thousands, or more) component parts E.g. fire into tiny flames Instead of animating each part by hand, provide rules and overall guidance for computer to construct animation
8
8cs426-winter-2008 When in doubt… Used to model particle-like stuff: dust, sparks, fireworks, leaves, flocks, water spray… Also phenomena with many DOF: fluids (water, mud, smoke, …), fire, explosions, hair, fur, grass, clothing, … Three things to consider: When and where particles start/end The rules that govern motion (and additional attached variables, e.g. colour) How to render the particles
9
9cs426-winter-2008 What is a particle? Most basic particle only has a position x Usually add other attributes, such as: Age Colour Radius Orientation Velocity v Mass m Temperature Type The sky is the limit - e.g. AI models of agent behaviour
Similar presentations
© 2024 SlidePlayer.com Inc.
All rights reserved.