Figure 3:

A generalization of a hyperspectral workflow. The workflow to extract information from sensor data and to bring it into a biological context to generate knowledge starts with data acquisition, hardware calibration, a proper normalization step, data pre-processing, and masking to focus on the object of interest—the plant—and to eliminate background (e.g., plant pot and stabilization sticks). Depending on the experiment set-up, data and the analysis type have to be divided into validation (val), training, and test dataset to train a model and then evaluate it on the test data. This is followed by the result interpretation and identification of diseases, stresses, or other properties of the plants. Vertical dashed lines describe in a general way the transition between the imaging process, the processing of the data, the generation of information, and interpretation of knowledge.