Study On The Biomass Estimation Model Of Camellia Based On Digital Image Processing

According to the disadvantage of traditional way, plant biomass information which is acquired by complexly and destructively. The article puts oil-tea camellia as the primary research object and the digital picture processing technology as research basis. It has adopted the digital picture processing technology discussing for three types of oil-tea camellia plantation region in different growth periods, different growth stage and biological image characteristic of section organs of different stages. The main content of work and consequence including;Firstly, according to the non-contact digital image and taking picture of the oil-tea camellia. Through digital picture processing software technology, extracting the oil-tea camellia information of side photo area, height, width, canopy height.Secondly, according to picture characteristics of the seedling stage, time of beginning to bear fruit, full productive age of the oil-tea camellia and analyzing dependency of different times of the oil-tea camellia and different characteristic variable and various combination variable. Choosing the highest coefficient of association as independent variable of the oil-tea camellia estimation model, there are side photo area, height, width, canopy height.Thirdly, putting single feature as independent variable; putting the oil-tea camellia as dependent variable. Through modeling approach of linear regression, steady linear regression, polynomial regression, random forest and support vector regression model, building different times of the oil-tea camellia estimation model and detecting the precision of estimation model. The results showed that young oil-tea Random Forest model estimation better, before the fruit of polynomial regression model to estimate better fruit period linear regression model to estimate better.Fourth, putting more characteristic datum as independent variable and building different times of oil-tea camellia estimation model separately. Through analyzing structure Vector regression, neural networks, stepwise regression, multiple linear regression, and Random Forest model and testing various model precision and comparing predictive effect of different modeling approach. As a result, during the three growth period of the oil-tea camellia, prediction accuracy of multiple linear regression was high and has a excellent model performance.