The automatic assembly planning with fasteners (AAPF) prototype system is described. Given a model of a product to be manufactured, the AAPF system produces a high-level assembly sequence for producing that product. The model is defined in terms of constructive solid-geometry primitives and nut, bolt, and screw fastener primitives. The system works backwards by disassembling the finished product model. Under the assumption that no internal or external forces are present, the reverse of a disassembly sequence is a valid assembly sequence. This assembly planning system pays attention to the individual characteristics of the fasteners to produce a more efficient assembly plan. By paying attention to the type, size, and location of each fastener, the fasteners can be grouped and ordered for assembly (where possible) according to similar fastener type, size, and location. Depending on the factory environment, this can reduce the need for moving subparts from work cell to work cell, retooling time, and the time spent at any given machine or work cell. Several examples, ranging from a simple blocks world model to a mechanical computer mouse, are given, including their run times.>

A fully autonomous computer-vision system needs the ability to automatically learn, or discover, from training views of objects the object information (models) needed for recognition. One approach to object recognition uses an evidence rulebase composed of objects with corresponding degrees of evidence for the various objects in the databases. A technique for automatically deriving the evidence rulebase from training views of objects is presented. This generated rulebase is shown to provide successful recognition of new views of those objects.>

An evidence-based recognition technique is defined that identifies 3-D objects by looking for their notable features. This technique makes use of an evidence rule base, which is a set of salient or evidence conditions with corresponding evidence weights for various objects in the database. A measure of similarity between the set of observed features and the set of evidence conditions for a given object in the database is used to determine the identity of an object in the scene or reject the object(s) in the scene as unknown. This procedure has polynomial time complexity and correctly identifies a variety of objects in both synthetic and real range images. A technique for automatically deriving the evidence rule base from training views of objects is shown to generate evidence conditions that successfully identify new views of those objects.>

We define and verify the utility of a pattern analysis procedure called sparse decomposition. This technique involves sequentially ``peeling'' sparse subsets of patterns from a pattern set, where sparse subsets are sets of patterns which possess a certain degree of regularity or compactness as measured by a compactness measure c. If this is repeated until all patterns are deleted, then the sequence of decomposition ``layers'' derived by this procedure provides a wealth of information from which inferences about the original pattern set may be made. A statistic P is derived from this information and is shown to be powerful in detecting clustering tendency for data in reasonably compact sampling windows. The test is applied to both synthetic and real data.

The recognition of objects in three-dimensional space is a desirable capability of a computer vision system. Range images, which directly measure 3-D surface coordinates of a scene, are well suited for this task. In this paper we report a procedure to detect connected planar, convex, and concave surfaces of 3-D objects. This is accomplished in three stages. The first stage segments the range image into ``surface patches'' by a square error criterion clustering algorithm using surface points and associated surface normals. The second stage classifies these patches as planar, convex, or concave based on a non-parametric statistical test for trend, curvature values, and eigenvalue analysis. In the final stage, boundaries between adjacent surface patches are classified as crease or noncrease edges, and this information is used to merge compatible patches to produce reasonable faces of the object(s). This procedure has been successfully applied to a large number of real and synthetic images, four of which we present in this paper.