Evaluation Of Garlic Growth Based On Convolution Neural Network

The integration of information technology and agricultural construction is not only an important starting point to promote the high-quality development of agriculture,but also an effective way to realize agricultural modernization.With the rapid development of modern computer technology,image recognition and classification technology based on artificial intelligence is widely used in crop growth monitoring,pest early warning,yield prediction and other fields.The world garlic looks at China.The informatization level of garlic is reflected in the research of price prediction,unit yield prediction,area prediction and public opinion analysis.However,there is a lack of monitoring and evaluation of garlic growth,resulting in the opacity of garlic growth information and restricting the development of garlic industry.To solve this problem,on the basis of literature analysis and field research,the research on garlic growth is carried out,the garlic growth evaluation service module is designed and developed,and the research results are applied to garlic growth monitoring.The specific research contents are as follows:（1）Identification and classification of garlic growing periodPreprocess the collected garlic image data,expand the data set by adjusting the image size,rotation,flip and adjusting the image brightness,remove the noise by means of color segmentation and foreground segmentation,extract the image features of each growth period of garlic by using convolution neural network model,and identify and classify the growth period of garlic according to these features.The results showed that the accuracy of convolution neural network model in identifying and classifying garlic seedling stage,differentiation stage and elongation stage were97.99%,96.14%and 96.92%respectively,and the comprehensive accuracy was97.02%.（2）Build garlic growth parameter evaluation modelAccording to principal component analysis and correlation analysis,the main growth parameters and image characteristics of garlic in each growth period were determined,and the aboveground biomass,leaf area index,plant height,canopy coverage and other indicators were selected.Convolutional neural network（CNN）,support vector machine（SVM）and random forest（RF）were used to estimate the growth parameters of garlic seedling stage,differentiation stage and elongation stage respectively.The linear regression model between the estimation data of growth parameters and the measured data was constructed.R~2and MSE were used to quantitatively evaluate and compare the estimation effects of the three models.The results show that the performance of convolution neural network model in estimating garlic growth parameters is better than support vector machine and random forest.（3）Evaluation and validation of garlic growthAccording to the estimation results of growth parameters of convolution neural network model,the evaluation system of garlic growth parameters is constructed,the growth parameters are standardized,and the scores obtained from the growth parameters are added to obtain the total score.There are four experimental areas a,B,C and D.the daily,growth periods and overall growth of garlic in the four experimental areas are evaluated and compared.It is found that the growth of garlic in experimental area a is the best and that in experimental area D is the worst.（4）Build garlic growth assessment service moduleAccording to the research results,the convolution neural network model is applied to growth monitoring.This module is divided into three parts:data query,data analysis and comprehensive service:the data query part shows the obtained environmental data,manual measurement data and real-time data of the Internet of things during the growth period of garlic;The data analysis section is the evaluation of garlic growth,input garlic growth pictures,and output the growth period and growth of garlic;The comprehensive service section includes a brief introduction to garlic,garlic processing technology,garlic preservation technology,etc.