Image Recognition Method Of Apple Leaf Diseases On Loess Plateau

Apple production has an important impact on the economy of Shaanxi and Gansu,which is one of the important pillar industries in the Loess Plateau.Due to the unique natural environment conditions,the quality of apples in the Loess Plateau region is very good.However,in recent years,due to natural disasters and improper cultivation and other reasons,the frequent occurrence of apple pests and diseases seriously affects the apple industry.In view of the frequent occurrence of apple diseases and lack of timely treatment of the problem,this paper compares the common apple diseases,such as alternaria mali roberts,mosaic and rust,investigates and studys them,collects the corresponding images as samples,analysises of image features,and begins preprocessing and feature extraction operation.Finally this paper develops the image recognition system of apple leaf development and identifies three kind of diseases.The main research contents are as follows:(1)According to apple leaf disease images in complex background,this paper studies the image preprocessing denoising and lesion segmentation problem,establishes the preprocessing process.Contrast enhancement was achieved by three linear stretch.using the improved median filter can effectively remove noise.By transforming the color model,the image is converted into L*a*b* color space.Using K-clustering algorithm separates foliage and background segmentation,and then using the maximum variance algorithm splits out the segmented lesion image.Simulation results show that this kind of lesion segmentation method can achieve better segmentation effect.(2)It studies on the extraction of effective features of apple disease images.The objects are tested from three aspects: color feature,shape feature and texture feature.It extracts H vector and draws the H-S histogram as the color feature of the lesion.It extracts shape feature according to the geometric characteristics of the lesion and invariant moment.Using gray level co-occurrence matrix analyzes texture characteristics.Finally it selects 13 parameters as classification features from the experiment of the 22 features.(3)It studies on the relevant research methods of pattern recognition and support vector machine model.Comparing with Bias decision method and artificial neural network methods from advantages and disadvantages,this paper selects the disease classification model based onsupport vector machine,designs the multi class support vector machine classifier model,determines the parameters of the model,and selects 13 features as classification of training.(4)This paper uses C# code to do the programming experiment,joins the Matlab program interface,and develops an identification system.The experimental results show that the classification method can identify apple diseases effectively,and can meet the needs of apple disease intelligent diagnosis.