Search NASA⌕ Search

Engineering topics

Vanderbrug, G. J.

Publications and source records attributed to Vanderbrug, G. J..

Edge and line detection in ERTS imagery: A comparative study

Several local edge detection operators were applied to a set of ERTS pictures of the Monterey, Calif. area. Gradient operators performed consistently better than laplacian operators in detecting edges. It was also found that if a grayscale normalization operation, histogram flattening, was applied to the pictures first, the edge detector outputs were greatly enhanced. The use of interpolation for more accurate location of edges on a digital picture was also briefly investigated. Curve detection operators were applied to the edge detector outputs; this had the effect of enhancing the edges while suppressing noise.

Eberlein, R.↗

Properties of heuristic search strategies

A directed graph is used to model the search space of a state space representation with single input operators, an AND/OR is used for problem reduction representations, and a theorem proving graph is used for state space representations with multiple input operators. These three graph models and heuristic strategies for searching them are surveyed. The completeness, admissibility, and optimality properties of search strategies which use the evaluation function f = (1 - omega)g = omega(h) are presented and interpreted using a representation of the search process in the plane. The use of multiple output operators to imply dependent successors, and thus obtain a formalism which includes all three types of representations, is discussed.

Vanderbrug, G. J.↗

Representations of the language recognition problem for a theorem prover

Two representations of the language recognition problem for a theorem prover in first order logic are presented and contrasted. One of the representations is based on the familiar method of generating sentential forms of the language, and the other is based on the Cocke parsing algorithm. An augmented theorem prover is described which permits recognition of recursive languages. The state-transformation method developed by Cordell Green to construct problem solutions in resolution-based systems can be used to obtain the parse tree. In particular, the end-order traversal of the parse tree is derived in one of the representations. An inference system, termed the cycle inference system, is defined which makes it possible for the theorem prover to model the method on which the representation is based. The general applicability of the cycle inference system to state space problems is discussed. Given an unsatisfiable set S, where each clause has at most one positive literal, it is shown that there exists an input proof. The clauses for the two representations satisfy these conditions, as do many state space problems.

Minker, J.↗