N-Dimensional Object Retrieval Using Multi-Level Decision Functions

The patent segments the collision detection process into multiple decision functions (zero-th through third decision functions) that operate at different levels of precision. The zero-th decision function performs rough filtering using minimum bounded boxes, while subsequent decision functions progressively refine the approximation accuracy for objects that pass through earlier filters, thus resolving the contradiction between speed and precision.

The patent applies local quality by using different approximation methods for different objects based on their characteristics. Objects that are clearly non-intersecting are handled with simple minimum bounded box approximations, while objects that require higher precision are processed with more accurate approximation methods, allowing the system to optimize computational resources locally rather than uniformly.

The patent segments the collision detection system into multiple decision functions with increasing complexity. The zero-th decision function uses simple minimum bounded boxes for all objects, while the first through third decision functions progressively apply more complex approximation methods only to objects that require higher precision, thus managing device complexity while improving measurement precision where needed.

The patent applies partial action by using enhanced approximation methods only for a subset of objects that require higher precision, rather than applying complex methods to all objects. This selective approach improves measurement precision for critical cases while keeping the overall system complexity manageable.

The patent makes the collision detection system universal by designing decision functions that can handle both simple rectangular queries and complex quadric surface queries. The same multi-level decision function framework adapts to different query types, maintaining high productivity for simple queries while gaining the adaptability to handle complex query shapes like shadows and circular areas.

The patent applies dynamics by making the query processing system adaptable to different query types. The decision functions dynamically adjust their processing based on the query shape, using efficient rectangular region processing for simple queries and more sophisticated quadric surface intersection detection for complex queries, thus balancing productivity and adaptability.

The patent applies preliminary action by performing rough collision filtering using minimum bounded boxes and the zero-th decision function before conducting detailed collision detection. This pre-filtering step eliminates obviously non-intersecting objects early in the process, so that detailed and time-consuming collision detection is performed only on a small subset of candidate objects, thus achieving high measurement precision without excessive processing time.

The patent maintains continuity of useful action by implementing a multi-level decision function system where each level builds upon the previous level. The zero-th decision function continuously filters candidates, and the first through third decision functions continuously refine the collision detection accuracy for remaining candidates, ensuring that computational resources are continuously applied to useful detection tasks rather than being wasted on obviously non-intersecting objects.