Research On Nondestructive Detection About Moisture Content Of Maize Leaf Based On Dielectric Properties

The moisture content is an important index in the healthy growth of crops,which effected the physiological metabolism and growth potential directly. It is of great significant to reach a nondestructive, fast and accurate purpose of detecting the moisture content of crop leaf. In recent years, it was widely used in non-destructive detection on moisture content of leaf based on dielectric properties. But during the experiment of measuring dielectric parameters of leaves, the traditional parallel electrode plate damaged the leaves with too much pressure or loosed the leaves with the little pressure, which effected detection accuracy. Duringdetecting moisture content of leaf, resistance, inductance and capacitance were commonly used, which were effected by the leaves shape and thickness. In this paper, the maize leaf was used as the research object. The LCR meter and a self-made parallel electrode plate, which could clamp the maize leaf, were used to measure the relative dielectric constantε′and dielectric loss factorε "of maize leaves. ε ′ and ε " were not effected by leaf shape and thickness. In this paper, a nondestructive method detecting moisture content of maize leaf based on the dielectric property technology was presented. A model was established to predict the moisture content of maize leaf in this paper. According to this model, an online detection system on moisture content of maize leaf was designed.The main research contents and conclusions are as following:(1) It was studied on the relationship among moisture content, frequency and dielectric parameters(the relative dielectric constantε′and dielectric loss factorε ")of maize leaf in this paper. It showed thatε′andε" decreased with the increasing frequency. Under the same frequency, ε ′ and ε "increased with the increasing moisture content of maize leaf. ε ′ and ε " were influenced in different degree by moisture content. Accordingly, the models were established to analysis the relationship between the moisture content and 3 kind of information variables(ε′、ε "and the combine ofε′andε").(2) In this paper, multivariate linear regression(MLR) and support vector regression(SVR) were used to establish the prediction models on moisture content ofmaize leaf. SWR and SPA, the characteristic variable selection methods, were used in this paper. SWR was used to optimize the MLR prediction model and SPA was used to optimize the SVR prediction model. In the result, the SPA_SVR prediction model owned the best performance in all moisture content of maize leaf prediction models,whose variables decreased from 72 to 10 through SPA, with coefficient of determination(RP2) in test set of 0.804 and root mean squared error(RMSEP) in test set of 0.0176.(3) In this paper, an online detection system on detecting the moisture content of maize leaf was developed based on the best prediction model above. The development environment was set up with Eclipse development platform and Tomcat server. The development programming languages included Java, JSP and CSS. My SQL database was used to store and read data. The system achieved some functions including the dielectric parameter input, the characteristic variable selection, the moisture content detection, the historical data query, which realized the on-line detection of the moisture content of maize leaf.