Multi-resolution description of the Rig

A rig structure is made of component nodes, where each node is an instance of a component class, with a given position and orientation in 3D space. Each component class is characterised by geometric, structural, and pictorial information at different levels of detail: a collection of descriptions of each component is maintained, such that each description is associated to a range of distances, and more detailed descriptions are used at closer distances. A rig is therefore encoded as a set of nodes that form a hierarchical structure. Each node provides multiple geometrical and pictorial descriptions of an object, as well as pointers to suitable recognition strategies, by instancing a specific component class at a given position in space. Each component class is defined by a list of Levels-of-Detail (LODs) which provide alternative descriptions of an object. Each LOD gives a single description, and it is related to a range of distances.

Beyond a certain distance, an object is considered invisible, both for rendering and for recognition, thus it has an empty description. In the middle ranges of distances, the object is perceived as a whole, hence it admits a more or less detailed description in terms of graphical primitives, and a specific recognition strategy typical of its component class. At close distance, the object is rather perceived as a complex of components, hence its structure must be detailed further in terms of its components, and specific recognition structures will be applied to such components. This last LOD, corresponding to a composite structure, is implemented as a list of links to instances of other component classes, thus giving the hierarchical structure of data.

Back to previous page