Object models include more than just shape and structure, and this section discusses the remaining information and its representation. Model invocation (Chapter 8) needs three types of object-linked information. (The work described below is by both myself ([67,72]) and Paechter .) The details and the use of this information is described in depth in Chapter 8.
The first type is interobject relationships, which provides indirect evidence.
Seven types of relationships are allowed:
subcomponent, supercomponent, supertype (family),
subclass (specialization), superclass (simplification),
arbitrary association and inhibiting.
The model also includes a weight to express the importance of the relationship.
The information is represented as:
relation OF objecttype1 IS objecttype2 weight
SUPERTYPE OF trash_can_outer_surface IS positive_cylinder 1.0
The second requirement is for constraints on feature properties (Chapter 6) specified by the acceptable value ranges and a contribution weight. Examples of this information are the expected areas of surfaces or angles at which surfaces meet. The value ranges are based on normal distributions about peak values, and may include or exclude the values near the peak. Additionally, some properties may just be required to be above (or below) the peak value. Altogether, six forms for property requirements are used.
The final invocation requirement is for subcomponent groups, which are lists of the immediate subcomponents seen from each distinct viewpoint.
Some of the extra information could have been derived from the geometric models (e.g. subcomponent relationships). For others, such as the importance of an attribute to the invocation of a model, the relationships or the methods for automatically extracting the information (i.e. for deciding on the key visible feature groups) are not well understood. Hence, the extra information is represented explicitly.