Recognition Of Apple Leaf Disease Basedon Extreme Learning Machine

With the development of information technology,image processing has become more and more intensive in many fields.At the same time,with the vigorous development of the apple industry and the continuous expansion of apple planting area,the impact of various diseases on apple growth has also increased,which seriously affects the yield and quality of apples and the economic income of fruit farmers in recent years.Studies have found that most of the pests and diseases will affect the growth of apple leaves.In addition,the shapes,colors and textures of the different diseases on the leaves are also different.The effective identification and prevention of diseases is one of the first problems to ensure the economic income of fruit farmers.However,the identification of apple leaf diseases mainly relies on the experience at present,and it is not ensure the timeliness and accuracy.Therefore,this paper applies image processing technology to preprocess disease on the images of apple leaves,extracts disease features,and then uses the extreme learning machine model to classify the features.The main research contents are as follows:(1)Image preprocessing of apple disease leaves.Aiming at the image features of apple leaf diseases,the image is denoised by median filtering and wavelet transform,which can preserve the image edge information.Then the histogram equalization technique is used to enhance the image.Finally,the threshold segmentation algorithm and the largest class are adopted.The OTSU segmentation algorithm performs image segmentation to achieve effective segmentation of lesions.(2)Feature extraction of apple leaf disease images.According to the difference of image disease characteristics,the color moment method is used to extract the color features of the lesions,the gray level co-occurrence matrix is used to extract the texture features of the lesions,and the Hu invariant moment method is used to extract the shape characteristics of the lesions,and finally twelve characteristics with significant differences such as the energy,correlation and roundness are selected as identification parameters.(3)Classification of leaf diseases.Applying the extreme learning machine model to realize the disease detection of apple leaves.Through the experimental comparison,the Sigmoid function is selected as the model activation function,and the number of hidden layer nodes is 600.The experiment shows that can achieve the best combination.(4)Design and test the apple leaf disease detection system.In Matlab2016 a and Visual Studio2013 environment,Matlab and C# mixed programming technology is used to realize apple disease detection system.The experimental results show that the system detection rate can reach 96.8%.