Real-time pattern recognition. II. Visual conjunctoid neural networks

For pt.I, see ibid., p.580-3 (1991). Keping Ma (1991) studied alternative approached to airplane recognition. In one comparative study, he evaluated various forms of conjunctoid and normal model performance, based on conjunctive features such as the number of acute, right, and obtuse angles appearing in the visual contour of an airplane. In this paper, the authors first review some Ma's related results. They then describe some of the mathematical details associated with the conjunctoid counterpart to the normal model. The authors also describe some preliminary steps toward automating the feature selection process for conjunctoid neural networks in airplane recognition settings, with an eye toward developing a quite general automatic data processing framework.<<ETX>>