Applying Constraint Satisfaction Techniques to 3D Camera Control

Abstract

Controlling an autonomous camera in three-dimensional (3D) virtual environments is a difficult task which manifests itself in many interactive computer graphics applications. Computer games [2] and guided exploration [1] are examples where autonomous cameras are required to provide suitable views of the scene as the user interacts with the environment. The camera system is expected to consistently provide a suitable view of the target object(s) for the user.

The camera control problem is often over-constrained. We successfully applied local search strategies to generate solutions to the problem in real-time. A cost is associated with each constraint (current constraints are: distance, height, occlusion, frame coherence rotation, and frame coherence distance), making some constraints more likely to be satisfied than others. A 3D game engine was developed to evaluate the camera’s real-time performance. Preliminary comparisons with related works indicate significant performance benefits from the use of local search.