Application Of Pattern Recognition Techniques In Plant Numerical Taxonomy And Chlorophyll Content Of Genus Camellia

Pattern recognition technology (PRT) is a integrated technology that combines of mathematics method and the computer technology to the internal rules of research objects and the analysis of the hidden nature, including cluster analysis, machine learning and many other kinds of methods. PRT has evolved considerably since20years ago and widely applied in various fields. In this study, the Clustering method (Clustering operated CV), Back propagation neural network (Back-propagation artificial neural networks, BP-ANN), Support vector machine (Support vector machine, SVM) and many other kinds of pattern recognition technologys are applied to Camellia plant for numerical taxonomy and chlorophyll content of nondestructive prediction.There are five main outstretched functions developed:First of all, numerical taxonomy and cladistic analysis of19species of Camellia L. were performed using floral morphology containing continuous and discrete units. The current study mostly supports the classifications of19species as proposed in previous works. In addition, it also agrees with combining the following species together:C. oleifera and C. vietnamensis; C. sasanqua and C. hiemalis; C. brevistyla and C. puniceiflora; and C. grijsii and C. shensiensis. Further, we propose that C. maliflora be recognized as avariety of C. sasanqua, and C. phaeoclada is best placed in Sect. Paracamellia. Moreover, we conclude that these species can be combined:C. tenii and C. miyagii; and C. confusa and C. fluviatilis. Our study indicates that the numerical taxonomy and cladistic analysis based on morphological characters of floral organ is useful in species classification, and this technique appreciated in Sect. Oleifera and Sect. Paracamellia can be used for identification and classification of other taxa.Secondly, Leaf characteristics provide many useful clues for taxonomy. We used back-propagation artificial neural network (BP-ANN) and C-support vector machines (SVM) to classify47species from three sections of genus Camellia (16from sect. Chrysanthae,16from sect. Tuberculata and15from sect. Paracamellia). The classification model was constructed based on seven leaf anatomy attributes including, area of adaxial epidermal cell (AAD), thickness of adaxial epidermal cell (TAD), thickness of palisade parenchyma (TPP), thickness of total leaf (TTL), thickness of spongy parenchyma (TSP), thickness of abaxial epidermal cell (TAB), and area of abaxial epidermal cell (AAB). Model parameters of C-SVM, comprising regularization parameter (C) and kernel parameter (y), were optimized by cross validation. The best classification accuracy of the three Camellia sections was achieved by the radial basis function (RBF) SVM classifier (with parameters C=32, γ=0.13) as well as the sigmoid SVM classifier (with parameters C=32, γ=0.13) that was up to84.00%in the training set and90.91%in the prediction set, respectively. Compared with BP-ANN, SVM yields slightly higher prediction accuracy, which indicates that it is feasible to accurately classify the three sections of Camellia using SVMs based on leaf anatomy data.Thirdly, the aim of this study was to develop a complementary approach to discriminate68species from five sections of Camellia (11from sect. Furfuracea,13from sect. Paracamellia,15from sect. Tuberculata,24from sect. Theopsis and5from sect. Camellia) using support vector machines based on fractal leaf parameters analysis (FA) and leaf red, green, and blue (RGB) intensity values. The results showed that the best classification accuracy was up to96.88%using the RBF SVM classifier (C=16, g=0.5). The linear kernel overall accuracy was90.63%, and the correct classification rates of40.63%and93.75%were achieved for the sigmoid SVM classifier (C=16, g=0.5) and the polynomial SVM classifier (C=16, g=0.5, d=2), respectively. A hierarchical dendrogram based on leaf FA and RGB intensity values was mostly on agreement with the generally accepted classification of the Camellia species. Therefore, SVM combined with FA and RGB may be used for rapidly and accurately classifying Camellia species and identifying unknown genotypes.Fourthly, traditional spectrophotometry is a destructive and time consuming method to measure leaf chlorophyll content. The aim of this study was to assess if a new system combining artificial neural networks (ANNs) with back-propagation algorithm and chlorophyll fluorescence (ChlF) can measure the leaf chlorophyll content in a rapid and nondestructive manner. Optimized ANN models were developed for predicting Chl a, Chl b, Chl a+b, and Chl a/b of60Camellia species from14sections, including total leaves, upper (sun-exposed) and lower (shaded-leaves). Seven ChlF parameters and diurnal variation of full sunlight were used as model’s input layer. The performances of ANN models were then tested against an independent data set. The correlation coefficients between experimental and model predicted values ranged from0.1549to0.8925in total leaves models. However, the prediction results are better when leaves are separated, especially in shaded-leaves models (0.9953-0.9983). ANNs-ChlF system was demonstrated to be a useful tool for nondestructively and rapidly assessing chlorophyll content of Camellia species, but can potentially be applied to any other green plants containing chlorophyll.Accordingly, the application of pattern recognition technology based on different organs like flowers and leaves of the basic anatomy, morphology and physiology data, which is feasible to Camellia plant accurate classification, especially RBF-SVM classifier, as the tool for plant taxonomy has potential application prospect, and it is worth for spreading application.