Construction Of A Diagnostic Model For Potassium Deficiency In Apple Leaves Based On Shape And Color Combination Features

Potassium is an indispensable nutrient element in apple tree growth and development,which plays an important role in promoting fruit yield,enhancing resistance,improving fruit flavor and so on.The potassium content in leaves was positively correlated with the potassium content in branches and shoots,which was the most direct organ reflecting the nutritional status of trees.Therefore,leaf nutritional analysis and diagnosis can be used as the basis for apple tree nutritional analysis and diagnosis.The common diagnosis methods of potassium deficiency in crops include empirical diagnosis,chemical determination,spectral analysis and image processing analysis.Empirical diagnosis is subjective and easy to lead to human error.The results of chemical determination method are accurate and reliable,but it has the problems of professional operation,poor timeliness and high risk coefficient.Spectral analysis is accurate,but it has many problems,such as high cost,large volume and inconvenient popularization.With the development and application of image processing technology in agriculture,image processing analysis and diagnosis method has the advantages of strong real-time,convenient and low cost,and provides methods and ideas for intelligent orchard management and supplementary topdressing information.This study took the leaves of apple trees in each growth period in 2021 as the research object,collected leaf images with a digital camera under natural lighting,extracted the color and shape features of the leaf images based on digital image processing technology,and established shape-color combinations through linear discriminant analysis.The characteristics of the apple tree leaf potassium deficiency diagnosis model based on the combination of shape and color characteristics were constructed,and the application system of the leaf potassium deficiency diagnosis model was designed.The main research work is as follows:（1）Leaf image preprocessing.In order to improve the accuracy and reliability of the image,Canon camera was used to collect the image of apple tree leaves during the whole growth cycle,and a series of image pre-processing work including image denoising,leaf segmentation,background removal,contour extraction,color change area extraction and image illumination removal were carried out,which laid a foundation for the subsequent feature extraction.（2）Extraction and optimization of leaf image color and shape features.First,digital image processing technology was used to extract the mean value of single-leaf R,G,B,H,S,V monochromatic components and the standard values of NRI,NGI,NBI,a total of 9 color features.Secondly,according to the physiological symptoms of potassium-deficient leaves of apple trees,6 basic shape features and 4 estimated shape features were added to form 19 color and shape features.Finally,through linear discriminant analysis,data dimensionality reduction and optimization were carried out on the color and shape features of leaf images,and the key shape and color combination feature factors of leaves in each growth period of apple trees were obtained.（3）Establishment of diagnostic model of potassium deficiency in apple tree leaves based on morphological and color combination features.Taking the shape color combination characteristic factor as the input parameter of the model,the potassium deficiency diagnosis models of apple leaves of LDA-SVM,LDA-RF and LAD-KNN were established respectively,and the optimal diagnosis model in each growth period was evaluated according to the four evaluation indexes of accuracy,accuracy,recall and F1-score The results showed that the diagnostic accuracy of LDA-SVM model was higher than that of LDA-RF and LDA-KNN models in flowering stage,young fruit stage,expanding fruit stage and maturity stage.（4）Application of diagnostic model of potassium deficiency in apple tree leaves.The validation sets of apple tree leaves in each growth period were collected and the image information of the leaves was saved.The established LDA-SVM leaf potassium deficiency diagnostic model was used for field experiments.The results verified the generalization ability and robustness of the model,and the pyqt5 toolkit of Python was used to complete the design and application of the apple tree leaf potassium deficiency diagnostic model system.