The aim of all image classification exercises is to assign each image pixel to a cover class. Various classification approaches have been developed that assign pixels in specific ways. The choice of approach depends on:
- the purpose of the classification (e.g., the type of features that are classified, their lifecycles),
- the characteristics of the image set (e.g. footprint, spectral, resolution, timing of acquisition), and
- the intended use of the classification results (e.g., visual display or extraction of statistics).
In this section, we discuss three classification approaches that are commonly used in agricultural applications. We take a look at their comparative advantages and disadvantages, and where applicable, provide examples stemming from the STARS project.