Study On Potato Disease Recognition Method Based On Fractional Differential Of Interpolation And FPCA

With the advancement of potato staple food strategy,potato has become one of China's important food crops.At present,the planting technology of patato is highly mature,and improving the yield and quality of potatoes is the focus of current research.However,in the process of potato growth,potato is prone to various diseases,resulting in a reduction in yield and quality,which causes a significant economic losses.Therefore,accurate and timely detection of diseases can effectively reduce economic losses.With the development of science and technology,studying the automatic identification methods of potato leaf diseases has important practical significance for improving the intelligence of crop disease diagnosis.The paper selects potato early blight and late blight as research objects,and combines image processing technology and pattern recognition technology to study the identification methods of potato early blight and late blight.The main work are as follows:(1)In order to improve the reliability of image segmentation,feature extraction and recognition,this paper proposes a interpolating operation based fractional differential mask(IOFDM).This algorithm can not only enhance the texture details of potato leaf images,but also remove noise.(2)In order to accurately extract the disease affected area of potato leaves,in this paper,the adaptive threshold method was used to extract the lesions,and the morphological processing method was used to perform subsequent processing on the extracted lesions.Finally,the segmentation effect is good.(3)Research on feature extraction and feature fusion method,and extracted 29 types of feature parameters.Main including: the means and variances of R,G,B,H,S,I,and V components of the HSV color model;means and variances of correlation,energy,moment of inertia and information entropy;7 hu invariant moments.By comparing the contribution rate of principal component analysis(PCA)and fractional principal component analysis(FPCA)to verify their fusion effect on the characteristic parameters of potato lesions.Experiments show that the FPCA is more suitable for dimension reduction of potato lesions.(4)On the MATLAB platform,using the BP neural network and support vector machines(SVM)to classify late blight and early blight of potato.The experimental results are as follows: IOFDM + FPCA + SVM classification accuracy of two diseases of potato is 98%.IOFDM + PCA + SVM classification accuracy of two diseases of potato is 88%.IOFDM + PCA + BP classification accuracy of two diseases of potato is 80%.IOFDM + SVM classification accuracy of two diseases of potato is 58%.IOFDM + BP classification accuracy of two diseases of potato is 54%.SVM classification accuracy of two diseases of potato is 52%.Therefore,this paper finally selected IOFDM+FPCA+SVM to classify early blight and late blight of potato,and the recognition accuracy was 98%.