Ground reference refers to the reference dataset for an accuracy assessment of a remote sensing classification. The process of obtaining ground reference is dedicated to support the production of suitable accuracy information. A sampling design (fitting to the produced image classification) determines the most appropriate distribution of sample locations (or regions). The response design consists of the evaluation protocol and the labeling protocol. The evaluation protocol initiates selecting the support region on the ground (represented by a pixel or polygon) where the ground information will be collected. Once the location and dimension of the sampling unit are defined, the labelling protocol is initiated and the sampling unit is assigned a hard or fuzzy ground reference label. This ground reference label (e.g. forest) is paired with the remote sensing-derived label (e.g., forest) for assignment in the error matrix.