A Synthetic Review of Various Dimensions of Non-Destructive Plant Stress Phenotyping

Abstract

Non-destructive plant stress phenotyping begins with traditional one-dimensional (1D) spectroscopy, followed by two-dimensional (2D) imaging, three-dimensional (3D) or even temporal-three-dimensional (T-3D), spectral-three-dimensional (S-3D), and temporal-spectral-three-dimensional (TS-3D) phenotyping, all of which are aimed at observing subtle changes in plants under stress. However, a comprehensive review that covers all these dimensional types of phenotyping, ordered in a spatial arrangement from 1D to 3D, as well as temporal and spectral dimensions, is lacking. In this review, we look back to the development of data-acquiring techniques for various dimensions of plant stress phenotyping (1D spectroscopy, 2D imaging, 3D phenotyping), as well as their corresponding data-analyzing pipelines (mathematical analysis, machine learning, or deep learning), and look forward to the trends and challenges of high-performance multi-dimension (integrated spatial, temporal, and spectral) phenotyping demands. We hope this article can serve as a reference for implementing various dimensions of non-destructive plant stress phenotyping.

Keywords: imaging; multi-dimension; plant stress phenotyping; spatial; spectral; spectroscopy; temporal.

Figure 1

Data-acquisition process for non-destructive optical-based plant stress phenotyping, involving 1D spectroscopy, 2D imaging, and 3D phenotyping. One-dimensional spectroscopy techniques include Visible-IR and ChlF spectroscopy; two-dimensional imaging techniques include visible, multispectral, hyperspectral, IR, and thermal-IR imaging; three-dimensional phenotyping techniques include LiDAR, X-ray CT, MRI, and PET. The corresponding acquired raw data and data type are also shown in this picture.

Figure 2

One-dimensional reflectance spectral data processing pipeline. After data collection, preprocessing is needed to remove the influence of irrelevant information and background noise in the results. Calibration of the model is used to find the correlation between the sample’s properties and absorbance, perform the fitness test of the model, and connect the attributes of samples with the preprocessed measured spectra. Validation of the model is used to predict the spectral signal of unknown samples based on the calibrated model and evaluate the model’s accuracy.

Figure 3

A typical chlorophyll fluorescence kinetic curve used to measure the leaf’s photochemical and non-photochemical parameters. A measuring beam means light that is too low to induce photosynthesis but high enough to elicit chlorophyll fluorescence. Actinic light means light is fit for photosynthetic function. Pulse means saturating flash that can transiently close all PS-II reaction centers, but the flash is short enough, so no increase in non-photochemical quenching occurs. (Reprinted with permission from Ref. [62]).

Figure 4

Two -dimensional image processing pipeline based on traditional machine learning (TML) and deep learning (DL). TML processing flow includes: data preprocessing, feature extraction, choosing the ML model, training the model, and finally obtaining the well-trained model (satisfying the accuracy). DL processing flow includes: preparing datasets and data preprocessing, choosing the DL model, training the model, and finally obtaining the well-trained model.

Figure 5

Three-dimensional phenotyping data processing pipeline. Data processing flow includes: data collection, data preprocessing, point cloud creation and organ segmentation, and derived traits extraction.