Abstract

Automatic structuring (feature coding and object recognition) of topographic data, such as that derived from air survey or raster scanning large­ scale paper maps, requires the classification of objects such as buildings, roads, rivers, fields and railways based on their shape. There is a considerable body of published work on the identification and classification of objects within images. Recognition is based on the matching of descriptions of shape. Several techniques have proved useful such as boundary chain encoding and moment invariants. The technique used here uses Fourier Descriptors. Based on a Fourier analysis technique applied to the boundary co­ordinates of an object expressed as complex numbers, Fourier descriptors are widely used in image processing to describe and classify shapes. The shape descriptors generated from the Fourier coefficients numerically describe shapes and can be normalised to make them independent of translation, scale and rotation. Classification is performed by comparing descriptors of the unknown object with those of a set of standard shapes, finding the closest match. Most applications using Fourier Descriptors deal with the classification of definite shapes, for example identifying a particular type of aircraft. To identify topographic objects the technique needs to be extended to deal with general classes of shape. Fourier descriptors are evaluated as general classifiers applied to broad classes of topographic shape (buildings, fields, roads etc.). To analyse their effectiveness, a corpus of shapes of classified objects was extracted from topographic large­scale digital maps. The descriptors of each shape were calculated and the results analysed. These indicate that normalised Fourier descriptors alone are unsuitable for such general classification. However, when applying the same Fourier method combined with other techniques it was found that they could help to discriminate between some classes of objects.