Studies On Diagnosis Of Camellia Nutrition Based On Computer Vision

Effects of nitrogen nutrition as the main nutritional components of crop quality, nitrogen nutrition status is vital to the whole growth period of crops. Nondestructive, accurate extraction of crop nitrogen nutrition status information not only can grasp the health of crops, but also it is an important prerequisite for the implementation of reasonable fertilization. The visual information reflected in crop nitrogen nutrition status will show through the leaf color, then, the computer vision technology can be obtained through the acquisition and analysis of the information of crop nitrogen status.According to cumbersome, complex and low efficiency of the traditional method in obtaining plant nitrogen nutrition content information and the camellia as the main research object base on the computer vision theory, correlation and inversion model of camellia blade between color characteristics and nitrogen nutrition was studied for establishing nondestructive, fast, accurate and automatic diagnosis system for prelimin ary basis. The main contents of paper include:(1)Nitrogen content and color variation of Camellia leaves. Through the computer vision technology to extract the characteristics of 98 kinds of leaf color value and SPAD value, according to the data obtained for the existence of variation. Based on the principle of maximum correlation, screened sensitive color features were significantly correlated with the SPAD value, the results obtained:G/L, Si, (G-B)/(G+B), b~*, (G-B)/(R+G+B), B-Y.(2)Nitrogen content and distribution characteristics of color sensitive. According to diff erent ages, different acquisition time and different parts of the three aspects of data acquisiti on, to find out the nitrogen content and sensitive color distribution features, to provide refere nee for modeling。(3)Build prediction model. Based on 108 sample data, select first 27 samples as model set, the remaining 81 sample data as verification set. Through the modeling method of linear regression, robust linear regression, polynomial regression, random forests and support vector regression, do research on modeling methods of prediction effect.