Research And Realization Of Pear Plant Phenotype Observation And Analysis System

Plant phenotype refers to all physical,physiological,biochemical characteristics and traits that reflect the entire process and results of plant structure,growth and development.It is mainly determined by genes and environmental factors,and it is embodied in plant type and physiological parameters,growth nodes,weed detection,disease detection and yield prediction,etc.Timely observation of plant phenotypes is of great significance in terms of crop safety and environmental sustainability,and plays an important role in guiding the flowering period,scientifically adjusting the amount of flowers,and detecting diseases and insect pests.Aiming at the problems of inaccurate,time-consuming and laborious flowering period prediction in traditional pear tree phenotype observation,this thesis carried out a research on the method of pear tree flowering period prediction based on PCA-BP neural network.Taking the pear tree as the research object,the meteorological observation data of Shijiazhuang Meteorological Station was analyzed by the principal component analysis method,and three principal components with a larger correlation coefficient with the flowering period of pear blossoms were obtained.The BP neural network model is introduced into the application of pear blossom forecasting,and the error is reduced to one day.The effectiveness of this model is verified by a comparative analysis with the traditional forecast model.Aiming at the problems of low efficiency and strong subjectivity in leaf disease identification in traditional pear tree phenotype observation,this thesis proposes an improved Bilinear-VGG16 network model.This model uses the VGG16 network structure with batch normalization layer added to the convolutional layer to extract the disease characteristics of the leaf image.Based on the traditional Bilinear-CNN,a bilinear pooling method with a single feature path is designed,which reduces the amount of parameters and calculations.At the same time,the two pooling methods of global average pooling and global maximum pooling are used to mine the fine-grained features of the image,which improves the defects of the traditional convolutional network model.Finally,the central loss function is added to form a joint loss function,which reduces the difference in the characteristics of the image,and the four-object classifier is used to further improve the performance of the network.After experimental verification and comparative analysis,the accuracy of the model is 96.7%,which is greatly improved compared to the traditional convolutional neural network model.Finally,based on the PCA-BP neural network predicting flowering model and the improved Bilinear-VGG16 disease recognition model,the pear tree plant phenotype observation and analysis system was developed.It realizes the effective prediction of the flowering period of pear trees and the rapid identification of leaf diseases,which is closer to daily applications.