Identification Of Drought Stress Status Of Tomato At Seedling Stage Based On Chlorophyll Fluorescence Imaging

Water is vital to crops and is an indispensable presence in the process of crop growth and development.Crop water deficit made crop water potential and turgor pressure lower than normal levels,causing crop metabolism disorders and photosynthesis termination,resulting in crop yield reduction or even death.Therefore,in view of the application requirements of crop drought stress status identification in agricultural refinement and unmanned management,this study took tomato at seedling stage as the research object,based on chlorophyll fluorescence imaging technology,to study the selection method of fluorescence parameters related to drought stress,and to mine important chlorophyll fluorescence parameters.At the same time,the chlorophyll fluorescence images were analyzed,and the image features that could characterize the drought stress state were extracted.Based on chlorophyll fluorescence parameters and fluorescence image features,a drought stress state identification model was established,in order to realize the monitoring of tomato drought stress at seedling stage and the determination of drought stress state,and to provide theoretical basis and technical guidance for the healthy growth of crops and rational irrigation.The main research and conclusions of this paper are as follows:（1）Drought stress test and data collection of seedling tomato.Taking tomato at seedling stage as the research object,a drought stress experiment of seedling stage tomato was designed.First,a pre-experiment was set up to collect the photosynthetic gas exchange parameters（net photosynthetic rate Pn,stomatal conductance Gs,transpiration rate Tr,intercellular CO₂concentration Ci）and chlorophyll fluorescence parameters（minimum fluorescence Fo after dark adaptation,maximum fluorescence Fm after dark adaptation,photochemical fluorescence quenching q P＿Lss in steady state light adaptation,quenching non-actinic light coefficient q N＿Lss in steady state,actual photon quantum efficiency at steady state time QY＿Lss）of the experimental group and the control group,respectively,and compared and analyzed the gas exchange parameters and common parameters of chlorophyll fluorescence to verify the effect of drought stress on tomato physiology at seedling stage.Finally,the chlorophyll fluorescence parameters and images of tomato plants on the 2nd,4th,6th,and 8th day of drought stress were collected to provide data support for subsequent research.（2）Chlorophyll fluorescence parameters are preferred.Among the 98 obtained chlorophyll fluorescence parameters,the fluorescence parameters that can characterize the drought stress state of tomato at seedling stage were screened.The necessity of parameter optimization was verified by normalization and correlation analysis of the fluorescence parameters.The fluorescence parameters were optimized based on the SPA,IRIV,and VISSA algorithms,respectively.The three algorithms select 12,29,and 25 chlorophyll fluorescence parameters,respectively.There were 5 parameters extracted by the three algorithms at the same time,which are the actual light quantum efficiency at moment L2 of the light adaptation process（QY＿L2）,the non-actinic fluorescence quenching of L3 during light adaptation（NPQ＿L3）,the photochemical quenching at time L2 in the light adaptation process of the Lake model（q L＿L2）,the steady-state light-adapted photochemical quenching of the Lake model（q L＿Lss）,and the photochemical quenching at time D3 dark relaxation process of the Lake model（q L＿D3）.This study combined the knowledge of plant photosynthetic physiology to analyze the changing trend of five public fluorescence parameters with different degrees of drought stress.（3）Chlorophyll fluorescence image feature extraction.The fluorescence images corresponding to the five common fluorescence parameters were analyzed,and the statistical features of histograms and the Gaussian fitting curve features of histograms were extracted.Pearson correlation analysis was performed between image features and drought stress status,and seven image features with strong correlation were obtained（correlation coefficient>0.6）,which are histogram mean of NPQ＿L3,q L＿Lss histogram mean and histogram Gaussian standard deviation,q L＿D3 histogram mean,the entropy mean of q L＿L3 gray level co-occurrence matrix,the mean moment of inertia of the QY＿L2 grayscale co-occurrence matrix,the large gradient advantage of the q L＿Lss grayscale-gradient co-occurrence matrix.It can provide support for the establishment of the next drought stress prediction model.（4）Establishment of a prediction model for tomato drought stress at seedling stage.Using three machine learning algorithms,LDA,SVM and KNN,based on the chlorophyll fluorescence parameters selected by the SPA,IRIV and VISSA algorithms and 7 chlorophyll fluorescence image features,the identification model of tomato drought stress state at seedling stage was established.The experimental results showed that among the models established based on chlorophyll fluorescence parameters,the LDA algorithm had the highest recognition accuracy,and the IRIV combined with the LDA algorithm had the best effect.The recognition rates for the 2nd,4th,6th,and 8th day of drought were 100%,95%,98%,and 98%,respectively.Among the models established based on chlorophyll fluorescence image features,the SVM algorithm has the highest recognition accuracy of 87.1%.Compared with the model established with only 5 public fluorescence parameters,the accuracy is increased by 3.4%.The recognition rates on the 4th,6th,and 8th day of drought were 90%,82%,87%,and 90%,respectively.