Some Extensions of the K-Means Algorithm for Image Segmentation and Pattern Classification

In this paper we present some extensions to the k-means algorithm for vector quantization that permit its efficient use in image segmentation and pattern classification tasks. It is shown that by introducing state variables that correspond to certain statistics of the dynamic behavior of the algorithm, it is possible to find the representative centers fo the lower dimensional maniforlds that define the boundaries between classes, for clouds of multi-dimensional, mult-class data; this permits one, for example, to find class boundaries directly from sparse data (e.g., in image segmentation tasks) or to efficiently place centers for pattern classification (e.g., with local Gaussian classifiers). The same state variables can be used to define algorithms for determining adaptively the optimal number of centers for clouds of data with space-varying density. Some examples of the applicatin of these extensions are also given.

Shape-Time Photography

We introduce a new method to describe, in a single image, changes in shape over time. We acquire both range and image information with a stationary stereo camera. From the pictures taken, we display a composite image consisting of the image data from the surface closest to the camera at every pixel. This reveals the 3-d relationships over time by easy-to-interpret occlusion relationships in the composite image. We call the composite a shape-time photograph. Small errors in depth measurements cause artifacts in the shape-time images. We correct most of these using a Markov network to estimate the most probable front surface, taking into account the depth measurements, their uncertainties, and layer continuity assumptions.

Quantitative Inference in a Mechanical Design Compiler

This paper presents the ideas underlying a program that takes as input a schematic of a mechanical or hydraulic power transmission system, plus specifications and a utility function, and returns catalog numbers from predefined catalogs for the optimal selection of components implementing the design. It thus provides the designer with a high level "language" in which to compose new designs, then performs some of the detailed design process for him. The program is based on a formalization of quantitative inferences about hierarchically organized sets of artifacts and operating conditions, which allows design compilation without the exhaustive enumeration of alternatives.

Automated Assembly Using Feature Localization

Automated assembly of mechanical devices is studies by researching methods of operating assembly equipment in a variable manner; that is, systems which may be configured to perform many different assembly operations are studied. The general parts assembly operation involves the removal of alignment errors within some tolerance and without damaging the parts. Two methods for eliminating alignment errors are discussed: a priori suppression and measurement and removal. Both methods are studied with the more novel measurement and removal technique being studied in greater detail. During the study of this technique, a fast and accurate six degree-of-freedom position sensor based on a light-stripe vision technique was developed. Specifications for the sensor were derived from an assembly-system error analysis. Studies on extracting accurate information from the sensor by optimally reducing redundant information, filtering quantization noise, and careful calibration procedures were performed. Prototype assembly systems for both error elimination techniques were implemented and used to assemble several products. The assembly system based on the a priori suppression technique uses a number of mechanical assembly tools and software systems which extend the capabilities of industrial robots. The need for the tools was determined through an assembly task analysis of several consumer and automotive products. The assembly system based on the measurement and removal technique used the six degree-of-freedom position sensor to measure part misalignments. Robot commands for aligning the parts were automatically calculated based on the sensor data and executed.

A Theory of Quantitative Inference Applied to a Mechanical Design Compiler

This thesis presents the ideas underlying a computer program that takes as input a schematic of a mechanical or hydraulic power transmission system, plus specifications and a utility function, and returns catalog numbers from predefined catalogs for the optimal selection of components implementing the design. Unlike programs for designing single components or systems, the program provides the designer with a high level "language" in which to compose new designs. It then performs some of the detailed design process. The process of "compilation" is based on a formalization of quantitative inferences about hierarchically organized sets of artifacts and operating conditions. This allows the design compilation without the exhaustive enumeration of alternatives.

3D Object Recognition: Symmetry and Virtual Views

Many 3D objects in the world around us are strongly constrained. For instance, not only cultural artifacts but also many natural objects are bilaterally symmetric. Thoretical arguments suggest and psychophysical experiments confirm that humans may be better in the recognition of symmetric objects. The hypothesis of symmetry-induced virtual views together with a network model that successfully accounts for human recognition of generic 3D objects leads to predictions that we have verified with psychophysical experiments.

Perceptually-based Comparison of Image Similarity Metrics

The image comparison operation ??sessing how well one image matches another ??rms a critical component of many image analysis systems and models of human visual processing. Two norms used commonly for this purpose are L1 and L2, which are specific instances of the Minkowski metric. However, there is often not a principled reason for selecting one norm over the other. One way to address this problem is by examining whether one metric better captures the perceptual notion of image similarity than the other. With this goal, we examined perceptual preferences for images retrieved on the basis of the L1 versus the L2 norm. These images were either small fragments without recognizable content, or larger patterns with recognizable content created via vector quantization. In both conditions the subjects showed a consistent preference for images matched using the L1 metric. These results suggest that, in the domain of natural images of the kind we have used, the L1 metric may better capture human notions of image similarity.

Learning Three-Dimensional Shape Models for Sketch Recognition

Artifacts made by humans, such as items of furniture and houses, exhibit an enormous amount of variability in shape. In this paper, we concentrate on models of the shapes of objects that are made up of fixed collections of sub-parts whose dimensions and spatial arrangement exhibit variation. Our goals are: to learn these models from data and to use them for recognition. Our emphasis is on learning and recognition from three-dimensional data, to test the basic shape-modeling methodology. In this paper we also demonstrate how to use models learned in three dimensions for recognition of two-dimensional sketches of objects.