287.

[Cover] [Contents] [Index]

Page 57

Figure 2.1 (a) Image spatial space, (b) Co-registration of three images, (c) Image feature space.

the original feature space. Orthogonal transforms, which accomplish this aim to a greater or lesser extent, are described later in this section.

In a survey of pattern recognition methodology in remote sensing, Landgrebe (1998) notes the need for accurate class statistics if supervised classifiers are to perform in a satisfactory way. The minimum size of a training data set depends to a considerable extent on the number of features used to characterise the objects to be classified. However, ground data are difficult and also expensive to procure. Landgrebe (1998) points out that the number of possible locations in feature space depends on the number of features and on the number of quantisation levels used for each feature. For example, the Landsat-7 ETM+ produces seven bands of data with 256

[Cover] [Contents] [Index]