Measurement Of Citrus Chlorophyll Content Inversion Model Based On Reflection Spectrum Data

As the most important pigment for photosynthesis and important part of synthetic organic nitrogen,chlorophyll intervene the overall course of plant growth,development and fruits.So how to quickly and accurately measure the content of chlorophyll provides important reference significance for watering and fertilizing the plant and assessing the nutrient level and the growth of plants.The citrus in a farm of Union city,California is the experiment samples in this article.The reflectance spectrum data of citrus leaves is rea-time taken by YY Labs Portable Spectrograph(OSVNIR-400/1100-S)without scathe.And measured value content of chlorophyll is gotten with spectrophotometric method.Then the feature spectrum parameters are extracted by analyzed the characters of the curve spectral reflectance and its differential curve.After that,the dimensionality of the feature spectrum parameters is reduced with principal component analysis.Then retrieval model of the content of chlorophyll is constituted and the reduced data of spectrum parameters is taken as the input of BP neural network.The main work and research results of this article are as following:(1)Getting the reflectance data of experimental samples with YYL Spectrum Portable Spectrograph.Smoothing the curve of the reflection spectrograms and relevant differential curves with mean filter and Savitzky-Golay filter,this can ensure the accuracy of the feature spectrum parameters.According to prior knowledge,15 feature spectrum parameters are taken as the relative factors of the content of chlorophyll.After Testing,all of the 15 feature spectrum parameters are correlated with the content of chlorophyll.(2)Reducing the feature spectrum parameters with PCA.Through calculation and analysis,the accumulative contribution rate of the first five principal components is 98.7673%.So the first five principal components are set as the inputs of BP neural network prediction model.(3)Retrieval model of content of chlorophyll is constituted based on BP neural network.The results of experiment show that BP neural network can well predict the content of chlorophyll in citrus leaves.Because of the connatural defects of BP neural network,Social Emotional Optimization(SEO)algorithm is introduced to update the link weights and threshold values of BP neural network with a new method in this article.The results of experiments show that the improved BP neural network by SEO performs better prediction accuracy in inverting the content of chlorophyll.The RMSE of the predict result of improve BP neural network is 0.0794,the MAPE is 5.31%.Compared with the traditional BP neural network,the RMSE and MAPE of improved BP neural network is reduced by 47.17%,2.92%,respectively.The experiment result shows that SEO algorithm can relive the inherent drawbacks of BP neural network to some extent.Compared with traditional BP neural network,the SEOBP model shows better prediction accuracy.This article laid the foundation for the measurement of plant chlorophyll content in agricultural production.