Research On Phenotype Acquisition And Analysis Method Of Tomato Fruit Based On Hyperspectral Imaging Technology

Tomato（Lycopersicon esculentum Mill.）is one of the most consumed and cultivated vegetables in the world and is very popular among consumers because of its nutritional value and variety of consumption options.High-throughput phenotyping is playing an increasingly important role in many areas of agriculture.Breeders will use it to obtain trait characteristics of crops so that they can select genetically valuable varieties,while growers will use it to obtain pre-harvest yield and quality profiles of crops in order to achieve accurate estimates of crop yield and quality.In most of the phenotypic analysis,image analysis plays an important role in phenotype acquisition studies and applications by virtue of its non-destructive nature and reduced reliance on human and material resources.In recent years,hyperspectral imaging has developed rapidly,particularly in the areas of component content traits and morphological and structural traits,and has provided a powerful technical tool for the development and application of automated phenotyping systems.Hyperspectral imaging（HSI）integrates the advantages of spectroscopy and imaging（spatial information）and has been successfully applied to the acquisition of component content phenotypes in tomato fruits.There are few studies on tomato fruit phenotype acquisition,and this paper proposed a non-destructive and rapid platform for determining the harvesting period,quality grading and transportation of tomatoes.The main research contents and conclusions are as follows:（1）Research on the tomato phenotype acquisition based on hyperspectral imaging technologyA hyperspectral imaging system was used to obtain spectral image information of tomatoes,and then the spectral images were first analyzed to obtain morphological and structural phenotypes（longitudinal diameter,transverse diameter,fruit shape index and weight）,followed by measurements of tomato color traits（L*,a*,b*,c*,h*,and a*/b*）and component content traits（firmness,SSC,lycopene,titratable acid,soluble sugar,and V_Cvalues）.Firstly,image processing was used to extract the transverse and longitudinal diameter information of the spectral images to obtain the morphological and structural traits of tomato fruits.It is the basis for the identification of tomato phenotypic information acquisition platform.Secondly,a qualitative model of tomato color was established based on the characteristic wavelengths selected by the successive projection algorithm（SPA）,and the results indicated that the accuracy of the LIBSVM classification model was 86.3%in the test set.Finally,partial least squares regression（PLSR）models for morphological structure and component content phenotype were established based on the characteristic wavelengths,in which the color phenotypes L*(~2_（1）=0.89、a*(~2_（1）=0.96、h*(~2_（1）=0.97and a*/b*(~2_（1）=0.96,the component content traits firmness(~2_（1）=0.74and lycopene(~2_（1）=0.72were better predicted.Thus,as a non-destructive detection tool,hyperspectral imaging can be used to predict tomato fruit quality and thus determine the optimal harvesting period,which in turn provides technical support for mechanized harvesting.（2）Research on the comprehensive phenotypes of tomato based on hyperspectral imaging technologyIn tomato quality detection,most of the detections are for single quality indexes and it is rarely possible to detect the overall quality of tomatoes at the same time.In this study,a visible and near-infrared（Vis-NIR）hyperspectral imaging system was used to collect images of three tomato varieties in order to search for comprehensive traits,and twelve tomato phenotypic traits（color and component content phenotypes）were used as reference standards.The trends and correlations of different phenotypic traits were analysed,and the comprehensive quality index（CQI）was proposed by factor analysis.CQI included a*,c*,h*,firmness,lycopene,V_Cand a*/b*.The best characteristic wavelengths were selected by SPA and a quantitative model was developed for the prediction of the comprehensive traits in tomatoes.The results indicated that the optimal model multiple linear regression（MLR）achieved good performance at~2₍（1）=0.87,RMSEV=1.33 and RPD=2.50.Spatial distribution maps of integrated phenotypic traits were generated based on the optimal model as a means of monitoring the integrated traits in tomato fruit.Based on hyperspectral imaging and chemometrics,the non-destructive prediction of tomato composite traits provides the basis for the development of a fruit phenotyping platform.（3）Research on a tomato phenotype analysis platform based on hyperspectral imaging technologyBased on the study of component content traits and morphological structure traits phenotypes at maturity stages,we laid the foundation for the design and study of an automated platform for tomato phenotype analysis.The tomato phenotype analysis platform included software design and hardware design,of which the hardware part was divided into image acquisition module,spectral acquisition module and control system.The software part consisted of a login interface,a self-test system,a greenhouse tomato phenotype acquisition system and a tomato phenotype acquisition system.The results indicated that the coefficients of variation for fruit diameter and coloring rate of the automatic tomato phenotype analysis system were 5%and 4%respectively,and the~2 of the component content trait fitting was over 0.70,indicating that the accurate tomato phenotype acquisition platform based on hyperspectral imaging technology is feasible and provides technical support for determining the harvesting period,quality grading and transportation of tomatoes.