Duda, R. O., & Munson, J. H. (1967). Graphical-data-processing research study and experimental investigation. STANFORD RESEARCH INST MENLO PARK CA.
The scope of the work under Phase I has been divided into five relatively independent tasks. These are out lined below.
This task involves the investigation of photographic processing techniques and their electronic equivalent for the simplification of input data. A study has been carried out on the use of unity gamma photographic transparencies for grey scale elimination leading to the suppression of fine detail and the creation of a solid outline around significant features of the landscape. The process may be of practical value for the simplification of aerial photographs before they are fed into a pattern recognition machine.
A survey will be made of electronic methods of performing pattern recognition. In Sec. III an outline is given of the characteristics of current techniques and the signature ” approach. This will be treated in detail in Report 2.
A small pattern recognition machine will be constructed capable of modification to various schemes of logic. A compatible set of components based on multi -aperture magnetic cores has been devised , and a general purpose machine is being constructed having an 8 – by- 8 retina 65 multi- level stores , and 5 binary outputs. The mechanical construction work is now 40% complete. Perceptron type logic will be used initially. A paper on the simulation of neural networks with multi-aperture magnetic cores is included.
Mathematical investigations are being undertaken. These are
divided into Parts (1) and (2), as outlined below.
Part (1 ) –The possible application of certain theorems of integral geometry is being investigated. These provide numerical relations governing such properties of geometrical figures as line length , area , and curvature , which are invariant under rotation and displacement.
Pqrt (2) — An investigation is being made of the logic of practical wiring systems for the improvement of the performance of machines of limited capacity, in specific situations.
Data storage and retrieval will be studied in relation to the special problems of non-alphanumeric graphical data processing. In Sec. VI , a brief outline is given of the chief areas of interest; a detailed account is deferred until Quarterly Progress Report 2.