Unsupervised machine learning via Hidden Markov Models for accurate clustering of plant stress levels based on imaged chlorophyll fluorescence profiles & their rate of change in time

Chlorophyll fluorescence (ChlF), a plant response in time to stressors, has long been known to be a useful tool to detect plant stress. Early and accurate plant stress detection is imperative in enabling timely and appropriate intervention. One major limitation of prior work is that, in general, only a few key inflection points of a localized section of a chlorophyll fluorescence signal are used to calculate single index values. These values yield very limited insight into stress level or type. In this paper, we present a method for plant stress classification that uses global (versus local) ChlF time-varying signal data acquired via imaging. We classify this time-varying-intensity-signal using a Hidden Markov Model (HMM). While HMMs have been used in other fields, in this paper we present their first application in the field of plant stress clustering and classification. We show how the proposed selection of a low-pass filtered plant's entire chlorophyll fluorescence signal profile, as a global feature selection, improves the accuracy of plant stress classification. Additionally, we show how the rate of change-in-time of the plant ChlF intensity time-varying profiles further improves the plant stress classification accuracy. Finally, we present experimental results to show the value and potential of the proposed method to enable more accurate and specific classification of plant stressor levels and stressor types.