Diagnosis Model Of Hydroponic Lettuce Deficiency Based On Machine Learning

The nutrient solution hydroponics model has developed rapidly in modern facility agriculture,and lettuce is one of the most widely grown crops in this model.In actual production,because the automatic management technology of nutrient solution is not perfect,hydroponic lettuce may show symptoms due to the lack of certain nutrients during the growth.Therefore,timely and accurate judgment of the health status of hydroponic lettuce growth is of great significance for its disease prevention and subsequent scientific management of nutrient solution.The characterization information of lettuce leaves can reflect its health status.In order to diagnose the types of nutrient deficiencies in hydroponic lettuce accurately,nondestructively,and quickly in the early stage of growth,this paper takes the canopy leaf images of the whole hydroponic lettuce as the research object,based on machine learning algorithms,to study and compare the recognition effects of different types of algorithm models,and propose an effective method for diagnosing deficiency..The main research contents and results are listed as follow:(1)Hydroponic lettuce deficiency cultivation experiment and image acquisition platform construction.In order to collect the nutrient-deficient samples of hydroponic lettuce required for the research,a nutrient-deficiency cultivation experiment of hydroponic lettuce was carried out.Based on Knop’s classical general hydroponic formula,3 experimental groups of potassium deficiency,calcium deficiency and nitrogen deficiency and 1 normal group were set up,totaling 4 groups.Real-time monitoring of the changes of parameters such as the temperature value,temperature value,p H value and EC value of the nutrient solution to ensure the stability of each environmental condition and the unity of the test condition variables;The leaf height,root length,etc.were monitored.The results showed that there were obvious growth differences between different types of hydroponic lettuce,which verified the feasibility of follow-up research and provided samples for the establishment of a hydroponic lettuce deficiency diagnosis model.(2)Feature extraction and dimensionality reduction of nutrient-deficient images of hydroponic lettuce.The collection and preliminary processing of the image samples of hydroponic lettuce deficient in nutrients were completed in this chapter.A PVC material image acquisition obscura was designed and constructed,and the Lifecam Studio high-definition camera was installed on the top of the image acquisition box,40 cm away from the cultivation tank.A total of 1440 image samples of the canopy leaves of hydroponic lettuce were collected.Preprocessing,comparing the different effects of the mean filtering and median filtering algorithms,completes image denoising,and further realizes image enhancement by using the method of histogram enhancement.The color image is processed based on the super green segmentation algorithm to obtain a grayscale image.and binary images;then through the extraction of color moments,grayscale co-occurrence matrix and boundary features,19 original feature information of color features,texture features and shape features of the image were extracted respectively,and finally based on principal component analysis method.Dimensionality reduction,a 6-dimensional new feature set is obtained to represent all the original features.(3)A machine learning-based diagnostic model for nutrient deficiencies in hydroponic lettuce.In order to realize the non-destructive,accurate and rapid judgment of the types of hydroponic lettuce deficiency,a diagnosis model of hydroponic lettuce deficiency was constructed based on the machine learning algorithm using image information.After feature extraction,1424 groups of valid samples out of 1440 groups of original samples were used for the study of the deficiency diagnosis model.In the Matlab 2019 a software environment under the Windows10 system,the C language mixed programming was used,based on the BP neural network algorithm and the support vector machine respectively.Algorithm and random forest algorithm to build models;after model optimization and effect comparison,the effect of the deficiency diagnosis model built by random forest algorithm is the best,and the recognition accuracy rate for the test set can reach 86.32%.The recognition accuracy of the test set is79.25% and 83.73%,respectively.(4)Design of a visualization system for the diagnosis of nutrient deficiency in hydroponic lettuce.In order to more intuitively show the results of hydroponic lettuce deficiency diagnosis to users,based on the research data in Chapters 3 and 4,using the Lab VIEW software platform,a visualization system for hydroponic lettuce deficiency diagnosis was designed.The system allows users to choose to view any image of hydroponic lettuce leaves collected within 24 hours,and display the image results of image preprocessing,the data results of feature extraction,and the final diagnosis results.After testing,the system runs stably and reliably.