Research On Image Segmentation And Recognition Of Apple Leaf Diseases

Purpose—Many plant diseases have brought misery to the world’s farmers in recent years,reducing global crop yields by an estimated 14 percent annually.Phytopathology aims to improve the survival chances of plants under adverse environmental conditions and parasitic microorganisms that can cause diseases.Temperature,PH,humidity and water are the environmental factors that affect the occurrence of plant diseases.Misdiagnosis of plant disease types can lead to the abuse of chemicals,on the one hand,resulting in economic losses,environmental imbalance and pollution,and on the other hand,the generation of drug-resistant pathogen strains,thus increasing the burden on farmers.At present,the disease diagnosis detected by human beings is time-consuming and expensive.The automatic segmentation and diagnosis of disease spots based on plant leaf images is more effective than the existing methods.Therefore,this paper will adopt deep learning method to complete the segmentation and recognition of apple leaf disease images.Method—Image acquisition,segmentation,preprocessing and feature extraction together consist of plant disease identification,and classification based on some models.Using a deep convolutional neural network model,apple plant diseases were detected from images of apple plant leaves and accurately classified into 6 categories.Classifications include "Healthy","Anthracnose Blight","Grey Spot","Rust","Spot Leaf Blight","Multiple Diseases".This paper improves the apple leaf disease dataset using data augmentation techniques,namely image rotation and flipping,histogram equalization,and color transformation.Based on the enhanced dataset,a new CNN network model was proposed for image segmentation of apple leaf diseases,and an improved VGG network model was proposed for apple leaf disease identification.Research results —The apple leaf disease image dataset is processed through data enhancement technology,and the proposed convolutional neural network model is improved through image segmentation technology;the image dataset segmented by convolutional neural network can be applied to train the optimized CNN for verification,training,testing.A higher CNN model,and on the basis of this model,an image recognition system for apple leaf disease is designed and implemented,so that the model has practical use value.Limitations of research — The types and quantities of images in the apple leaf disease image dataset are quantitative.The improved model can only detect the types of apple leaf diseases provided within the research range,which can only represent the most common diseases of apple leaves.It is uncertain whether the leaf disease species outside the dataset can be identified;If the background of the image is more complex,whether it can be segmented and recognized accurately needs to be verified.Actual impact—Compared with traditional methods,applying deep learning to image segmentation and recognition of apple leaf disease can not only improve the accuracy of segmentation and recognition,but also simplify the corresponding segmentation and recognition process,obtain a better model,and solve the problem of fruit farmers in planting.Prevent the problems of high work intensity,long time consumption and low efficiency when fruiting.Originality —The apple leaf disease image data set is processed by data enhancement technology,and effective segmentation is achieved through feature extraction to improve the utilization of image lesions;the improved CNN model is used to identify the segmented images,and finally a higher recognition accuracy rate is obtained.Model.