Knowledge about required functionality of an object can be used as an effective representation for a generic object category (e.g. "chair", "cup", or "hammer"). This approach to object representation and recognition has recently become an active area of research. We explore a scenario in which a robot senses the environment to obtain an initial partial shape model of an object. If the information in this initial model is not sufficient to hypothesize a possible function for the object, then additional view(s) may be suggested. Once a possible function is hypothesized, a plan is formulated for interacting with the object to confirm that its material properties are compatible with the hypothesized function. The module for reasoning about partial shape models has been evaluated on over 200 shape models acquired from range images. The module for carrying out a function verification plan has been evaluated in a simulated environment using the ThingWorld (TW) system.

Functionality-based recognition systems recognize objects at the category level by reasoning about how well the objects support the expected function. Such systems naturally associate a ``measure of goodness'' or ``membership value'' with a recognized object. This measure of goodness is the result of combining individual measures, or membership values, from potentially many primitive evaluations of different properties of the object's shape. A membership function is used to compute the membership value when evaluating a primitive of a particular physical property of an object. In previous versions of a recognition system known as Gruff, the membership function for each of the primitive evaluations was hand-crafted by the system designer. In this paper, we provide a learning component for the Gruff system, called Omlet, that automatically learns membership functions given a set of example objects labeled with their desired category measure. The learning algorithm is generally applicable to any problem in which low-level membership values are combined through an and-or tree structure to give a final overall membership value.

Recent work in computer vision has demonstrated positive results in reasoning about possible object function based on analysis of only the object shape. While shape properties are important, verification of actual functionality generally requires consideration of properties beyond pure static shape. In particular, dynamic physical properties such as the degree of deformation or rigidity under applied force are essential to the function of many objects. The work described in this paper combines reasoning about object shape with reasoning about deformation under applied forces for recognizing (categorizing) an object according to its function. Initial reasoning about the static 3-D object shape provides a function verification plan. This plan is a sequence of interactions that lead to confirming or rejecting the suggested object function. The interactions involve applying test forces to elements of the object structure as identified in the reasoning about static 3-D shape.>

Previous work on the recognition of objects by reasoning about their functionality has not dealt with objects that have moving parts. In this paper we introduce a scenario in which object recognition is accomplished by first deriving an articulated shape model from an observed sequence of 3-D shapes and by then reasoning about the possible functionality of the articulated shape model.

Representation schemes traditionally used in model-based vision are contrasted with the “function-based” representation scheme. A system which utilizes function-based representation has been implemented and tested, using the object category “chair” for case study. Function-based description is used to recognize classes and identify subclasses of known categories of objects, even if the specific object has never been encountered previously. Interpretation of the functionality of an object is accomplished through qualitative reasoning about its 3-D shape. During the recognition process, evidence is gathered as to how well the functional requirements are satisfied by the input shape. An investigation of different types of operators used in the combination of the functional evidence has been made. Three pairs of conjunctive and disjunctive operators have been used in the recognition process of more than 100 object shapes. The results are compared and differences are discussed.

The authors report the implementation and evaluation of a function-based recognition system that takes an uninterrupted 3-D object shape as its input and reasons to determine if the object belongs to the superordinate category furniture, and if so, into which (sub)category it falls. The system has analyzed over 250 input objects, and the results largely agree with intuitive human interpretation of the objects. The study confirms that a relatively small number of knowledge primitives may be used as the basis for defining a relatively broad range of object categories. The greatest derivation from intuitive human interpretation occurs with objects that humans would not typically label as one of the known categories defined, but which have some novel orientation in which they could serve the function of one of these categories. This is because the system uses a purely function-based definition of the object category.>

The work which demonstrates the feasibility of a different approach to 3-D object recognition is described. The authors construct a definition of a generic object category, such as a chair, in terms of the function required of the object. This definition is based on qualitative reasoning about 3-D shape, and does not imply any particular geometric or structural model for an object. Thus, this approach has the potential to lead to recognition systems of much greater generality than current CAD-based or model-based approaches.>

An attempt is made to demonstrate the feasibility of defining an object category in terms of the functional properties shared by all objects in the category. This form of representation should allow much greater generality. A complete system has been implemented that takes the boundary surface description of a 3D object as its input and attempts to recognize whether the object belongs to the category 'chair' and, if so, into which subcategory if falls. This is, to the authors' knowledge, the first implemented system to explore the use of a purely function-based definition of an object category (that is, no explicit geometric or structural model) to recognize 3D objects. System competence has been evaluated on a database of over 100 objects, and the results largely agree with human interpretation of the objects.>

Several researchers have previously described approaches to 3-D object recognition which use nonlinear optimization to control the matching of features of a 3-D object niodel to features found in an image. Recognition, in this context, includes estimating the parameters of translation and orientation of the object. A major problem acknowledged by previous researchers is how to efficiently choose a set of starting paranleter estimates which will avoid recognition errors due to local minima. The unique contribution of this paper is that it outlines an approach for using the perspective projection aspect gruph representation to alleviate the problems encountered by previous researchers, describes a particular implementation of this general approach, and presents data to illustrate the effectiveness (of the approac!].