International Journal of Pattern Recognition and Artificial Intelligence

This paper introduces a syntactic omni-font character recognition system. The "omni-font" attribute reflects the wide range of fonts that fall within the class of characters that can be recognized. This includes hand-printed characters as well.

A structural pattern-matching approach is employed. Essentially, a set of loosely constrained rules specify pattern components and their interrelationships. The robustness of the system is derived from the orthogonal set of pattern descriptors, location functions, and the manner in which they are combined to exploit the topological structure of characters.

By virtue of the new pattern description language, PDL, developed in this paper, the user may easily write rules to define new patterns for the system to recognize. The system also features scale-invariance and user-definable sensitivity to tilt orientation.

This paper describes pattern recognition techniques for discriminating between earthquakes and underground nuclear detonations from teleseismic signals. The features investigated are obtained by Maximum Entropy Spectral Analysis techniques. Classification performance is evaluated on a seismic database recorded at the Large Aperture Seismic Array (LASA) in Montana. The events are randomly divided into design and test subsets. Several pattern recognition algorithms including Baye's quadratic classifier, the minimum distance classifier and the nearest neighbor rule are applied. The best results are obtained with spectral ratios from windows taken at regular intervals within the coda combined with bodywave magnitude. With the Baye's classifier, 85% correct classification was achieved for the test set events. For a reduced dataset containing only central Asian events, 96% was achieved.

Machine vision systems used in industrial applications must execute their algorithms in real time to perform such tasks as inspecting a wire bond or guiding a robot to install a part on a car body moving along a conveyer. The real time speed is achieved by employing simple-minded algorithms and by designing parallel architectures and parallel algorithms for some tasks. The majority of the work on parallel architectures has been limited to architectures that support image processing, but not mid- or high-level vision In order for more complex vision algorithms to execute in real time, a more flexible architecture is needed.

Our conceptual approach to the problem is a reconfigurable computational network. Each configuration of the network implements an algorithm or class of algorithms A high-level language expresses the algorithms in a relational form that can be easily translated to the specification for a configuration. The language must be able to encode low-, mid-, and high-level vision algorithms and to efficiently handle not only pixel data, but also higher level structures. In this paper we describe a dataflow language called INSIGHT, which we have designed to meet these needs, and give several examples of parallel machine vision algorithms expressed in the language.

We present a new approach to the segmentation problem by optimizing a criterion which estimates the quality of a segmentation. We use a graph-based description of a partition of an image and a merging strategy based on the optimal use of a sequence of criteria. This method separates the strategy of making use of the segmentation criteria from their definition. An efficient data structure enables our implementation to have a low algorithmic complexity. Our method offers a general framework for solving a large class of segmentation problems. We show how to adapt this method to segment 2-D natural images including color images. This algorithm is also used for segmentation of 3-D images.

In this paper, we give purity-preserving conditions for iconic operators, and present a formal definition of degree of icon purity by extending an icon system to a fuzzy icon system, based upon the theory of fuzzy sets. We also show an icon system can be modeled by means of an attribute grammar.

This paper describes EJAUNDICE, which is designed to be a general-purpose expert system building tool. Considerations behind a number of design decisions for purposes of generality are examined. EJAUNDICE provides several control schemes, including biphasical control with goal-directed reasoning, data-driven processing, and control blocks, and integrates rule-based, frame-based, and logic-based reasoning paradigms in its framework.

The issues of knowledge representation and input/output in developing a Chinese expert system are also investigated.

In the case of expert systems, when premises from two different rules argue for the same conclusion, there exist a number of methods for combining the certainties in the conclusion, obtained from each premise individually, to yield a total certainty in the conclusion. When these certainties are interpreted as probability values, it is necessary in all these methods to make assumptions of independence concerning the premises and conclusion. It is proven here that there does not exist a procedure which yields, for all probability functions, meaningful bounds for the probability of the conclusion based on two premises in terms of the probabilities of the conclusion based on each premise individually. Additionally, it is proven that there does not even exist a procedure which yields, for all probability functions, meaningful bounds for the probability of the conclusion based on two premises in terms of the probabilities of all one- and two-member combinations of the premises and the conclusion.

Finally, suggestions are made as to when to make assumptions of independence and when to take other alternatives.